Interim Analysis: What Regulatory Affairs Professionals Need to Know

Transcription

1 r e g u l a t o r y a f f a i r s 703 Interim Analysis: What Regulatory Affairs Professionals Need to Know David Coutant, PhD, RAC Principal Research Scientist, Eli Lilly and Co, Indianapolis, Indiana A data monitoring committee (DMC) is often responsible for monitoring a clinical trial via interim analyses. Those responsible for regulatory strategy and implementation need to be fully engaged with this oversight process to ensure that the trial will achieve its goals with regulators. In the United States, both 21 CFR and 21 CFR describe sponsor responsibilities related to ongoing clinical trials conducted under an Investigational New Drug Application. Herein, I review the types of trials that may benefit from DMC involvement and provide an example of a DMC clinical trial model. I highlight important features of the roles of the sponsor, DMC, statistician, and the regulatory professional in interim analysis. I review the importance of early planning, knowing when to release interim data, and avoiding type I error rate inflation. Finally, I examine key points that lead to early stopping or study alteration. Key Words Interim analysis; Data monitoring committee; DMC; Data and safety monitoring board; DSMB Correspondence Address David Coutant, PhD, Eli Lilly and Company, Lilly Corporate Center, Indianapolis, IN ( cout@lilly.com). I N T R O D U C T I O N There exist many different models for clinical trial (CT) monitoring and for interim analyses. Most commonly, monitoring is done by a data monitoring committee (DMC), particularly in the case of large multicenter phase 3 trials. The DMC has many different names; it is also known as a data and safety monitoring board or a data and safety monitoring committee. The purpose of this article is to provide advice on best practices to the regulatory professional (RP) involved with DMCs to map out the path to success and to point out pitfalls to avoid. For further understanding of regulatory considerations with regards to DMCs, readers are encouraged to familiarize themselves with available regulatory guidances (1 5). Note that the roles and functions of DMCs are distinct from the functions of the investigative review board (IRB), also known in the European Union (EU) as the ethics committee. The ethics committee (5) is responsible for ensuring the protection of the rights, safety, and well-being of human subjects involved in a particular clinical trial. In comparison, the DMC is made up of a small committee of independent experts whose responsibility is to periodically assess ongoing clinical trial data for evidence of treatment benefit or possible harm. The IRB does not have access to unblinded data, as does the DMC. Both groups, however, do have contributing responsibilities to ensure the rights, safety, and well-being of trial subjects. Next I discuss the roles of sponsors, the DMC, statisticians, and the regulatory affairs professional involved with interim analysis (and DMCs). I review the importance of careful initial study planning, releasing interim data at the right time (if at all), and avoiding type I error rate inflation. Finally, I examine interim analyses that lead to early stopping (for success or futility), study alteration, and study enrichment. O V E R V I E W O F I N T E R I M A N A LY S I S R O L E S A N D R E S P O N S I B I L I T I E S ROLES OF SPONSOR When the situation calls for use of a DMC, it is generally ill advised for the sponsor to have access to unblinded interim data. So doing can present substantial risk to the integrity of the trial and can lead either to further unblinding (eg, either of study participants or investigators) or increased bias in the trial results (1). For these reasons, although small companies might be tempted to include the CEO or CFO among those that are unblinded to the data, in the author s opinion this practice should be strongly discouraged, lest a CT be perceived as Drug Information Journal, Vol. 44, pp , /2010 Printed in the USA. All rights reserved. Copyright 2010 Drug Information Association, Inc. Submitted for publication: December 23, 2009 Accepted for publication: May 27, DIJ 44(6) indd 703 9/7/2010 4:25:51 PM

2 704 r e g u l a t o r y a f f a i r s Coutant F i g u r e 1 DMC Flowchart of Statistical Interim Analysis Review Institutional review board or ethics committee provides ethical oversight Clinical centers investigators conduct trial Central labs collect PK, PD and patient demographic data Data management center (sponsor or CRO) assembles and organizes data for statistical analysis Adjudication committee assesses & adjudicates endpoints. Independent adjudication not needed for all trial types. Steering committee or executive board oversees clinical study conduct, and adjusts study protocol when justified Sponsor designs investigative clinical trial - Sponsor Statistician - Sponsor Senior Mgt - Sponsor Safety Committee Statistical analysis center independent statistician provides interim analysis Independent DMC closed review of unblinded interim data Legend: Parties blinded vs. those unblinded to interim results Blinded Unblinded No Does DMC recommend major changes (eg, to primary endpoint) No Does DMC recommend stopping early? Yes Yes No Does sponsor internal review committee agree with DMC? No Does sponsor internal review committee agree with DMC? Yes Inform regulatory agencies Yes Inform regulatory agencies Stop trial early Discuss with DMC to address differences of opinion compromised to an extent that it would not be accepted by regulatory authorities. Note that in some cases, such as in open-label trials with special concerns about safety, there may be good reason for the sponsor to access the unblinded interim data (1). Key sponsor representatives might also need to look at the unblinded interim data to confirm recommendations by the DMC for early trial termination. This possibility is illustrated in Figure 1 by the presence of an internal review committee that would include a handful of subject matter experts that are neither part of the sponsor s steering committee nor part of the DMC. Any sponsor employee that looks at the unblinded data should avoid any subsequent involvement in trial conduct and should minimize or eliminate interaction with study team members, or risk adversely 06 DIJ 44(6) indd 704 9/7/2010 4:25:52 PM

3 Interim Analysis r e g u l a t o r y a f f a i r s 705 affecting the integrity of the trial. Note that whenever there is a chance that the sponsor might need to look at unblinded interim data, it is imperative that criteria for such access should be established in the protocol and understood by the DMC prior to study start (1). Further, before the sponsors unblind themselves to the interim data, it is recommended that they first consult with the relevant regulatory agency. ROLES OF DMC A clinical trial DMC is a group of individuals with pertinent expertise that reviews on a regular basis accumulating data from one or more ongoing clinical trials. The DMC advises the sponsor regarding the continuing safety of trial subjects and those yet to be recruited to the trial, as well as the continuing safety and scientific merit of the trial (1). DMCs hold both open and closed sessions. The structure of these sessions is outlined in both the 2006 FDA Guidance (1) and the 2005 WHO guideline (2) on DMCs. At the open sessions, the sponsor s study team, the steering committee, clinical investigators, and, if necessary, independent consultants are invited. These sessions focus on reviewing study conduct and progress, and include review of blinded, pooled efficacy and safety data. In contrast, attendance at the closed sessions is restricted to DMC members, and includes review of the unblinded interim data. The DMC should be guided by accommodation for type I error (ie, chance of false positive assessment) and carefully thought-out stopping rules (eg, stopping for futility or for overwhelming efficacy) that were established a priori. Note that if the sponsor is aware of important safety concerns prior to study start, it is suggested that the DMC should be structured so that it can act quickly, if urgently needed. Alternatively, pauses in study conduct can be built in to allow time to review and act on interim data. Following its unblinded interim analysis, the DMC will issue a terse written statement to the sponsor. This typically includes a recommendation to: (a) continue the trial unchanged; (b) modify protocol, (c) suspend or pause study; or (d) terminate early. DMC involvement may be merited in several situations (1 3,6), as outlined in Table 1. The FDA and the European Medicines Agency (EMA) have slightly different stances on what trials are most appropriate for the inclusion of DMCs. The FDA will consider interim analysis data from phase 1b and phase 2a studies (the learn phase of development) as well as phase 2b and phase 3 (confirmatory) trials, but believes that issues of unblinding and statistical interpretation of interim analysis are of greater impact in confirmatory trials. Specifically, in their 2006 Guidance (1), the FDA mentions that certain DMC-recommended trial adjustments can be merited if driven by the intent to reduce the occurrence of certain emerging adverse events. However, the same guidance (1) mentions that DMC-recommended changes to efficacy endpoints or analysis subgroups is not recommended due to the DMC being unblinded to the data and the putative damage to data interpretability that may ensue. In contrast, the 2005 EMA guidance (3) indicates simply that if modifications to confirmatory trials are to be allowed (eg, increasing sample size), they need to be planned for in advance in both the study protocol and the DMC charter. Although monitoring study progress may not be one of the specified roles of the DMC, sluggish study enrollment and poor adherence to desired study conduct can be serious impediments to a successful study. In such cases it is appropriate for the DMC to meet with the steering committee to cooperatively recommend adjustments (7). Interim reviews by the DMC should also include evaluation and consideration of external factors, to safeguard the credibility of final study results. Lengthy trials can be affected over time by evolution in the understanding of the disease, changes in the defining characteristics of the population of interest, or baseline drift in standard-of-care treatment. These external changes may lead to the need to modify certain trial aspects. ROLES OF STATISTICIANS Important questions to consider are how many statisticians need to be involved in a clinical trial with DMC oversight, and how the role of Drug Information Journal 06 DIJ 44(6) indd 705 9/7/2010 4:25:52 PM

4 706 r e g u l a t o r y a f f a i r s Coutant T A B L E 1 Situations That Strongly Suggest Use of DMC Controlled trials with mortality and/or severe morbidity as a major endpoint. Could envision early stopping in case of highly significant positive response. Controlled trial of a new molecule entity (NME) that is intended to reduce mortality or severe morbidity. Could envision early stopping in case of highly significant positive response. Controlled trials where interim analysis data might call for early termination for safety reasons, for instance where adverse events or lack of effect could lead to increased mortality or morbidity. Any emergency trial where informed consent is waived (required by regulation). Adaptive design for pivotal registration trial with preplanned interim analysis. When Are Data Monitoring Committees Needed? Situations Where DMC Might Be Useful Long-term symptom trials (eg, rheumatoid arthritis), where stopping for futility would be important. Trials of long duration where the design or data analysis is complex, particularly where interim data may lead to questions on participant safety or may lead to alterations in study parameters (eg, endpoints, size). Phase 1 or phase 2 trials where risk is very high, for example, for first-in-class NME or NME with previously established significant safety risks (eg, drug class with history of adverse events such as antiarrhythmics). A DMC could provide credible independent external oversight and enhance patient safety. Note that blinding of interim data is not typically an issue in this situation. Trials that involve particularly vulnerable populations (eg, children or mentally handicapped) even for noncritical indication. Clinical trials for NMEs that have likely chance of resulting in adverse events of serious harm, albeit not mortality or severe morbidity. Trials of very short duration, but that involve high risk of adverse events such that special accommodations are made for the DMC to have quick access to review and act on interim data. each statistician is defined in these situations (8). In 2004, DeMets and Fleming (9) outlined four different roles of a statistician in interim analysis, including (a) the DMC statistician, (b) the independent statistician from a statistical analysis center (SAC); (c) the statistician from the sponsor s study team, and (d) the statistician on the Steering or Executive Committee (see Figure 1 to view how these groups might interact in a trial involving a DMC). DeMets and Fleming (9) indicated that the DMC statistician should not be a sponsor employee, should have no direct involvement in the trial, and should be unblinded to the interim data. The presence of the independent SAC statistician allows the DMC to see the statistical analysis of the interim analysis data while keeping the sponsor and the clinical research organization (CRO) blinded (10). The statistician on the sponsor s study team is responsible for input into the initial study, and possibly involved with the analysis of the final study data, or, using dummy randomization codes, creating the interim analysis programming to be used by the SAC. During initial design of the statistical analysis plan (SAP), the sponsor s statistician needs to thoroughly consider and address all sources of bias (eg, who is to be blinded versus unblinded, how results will be disseminated, etc). The statistician on the steering or executive committee will not be directly involved with the conduct of the trial, but will be responsible for deciding on a course of action following the DMC s recommendation. Note that a relationship diagram for the conduct of an industry-sponsored clinical trial involving an SAC and DMC was first described in 2001 (11); the flowchart in Figure 1 builds on this idea and describes it in greater detail. Although the use of four different statisticians would be the safest in ensuring blinding 06 DIJ 44(6) indd 706 9/7/2010 4:25:52 PM

5 Interim Analysis r e g u l a t o r y a f f a i r s 707 and trial integrity, other possibilities exist. For instance, roles a and b could be combined and taken care of by a single statistician, while roles c and d could also be combined. It is best to keep separate the blinded and unblinded roles, if at all possible, in the author s opinion. Note that if a sponsor statistician has access to the unblinded data, and is in close proximity to colleagues during study conduct, facial expressions or other nonverbal clues may inadvertently transmit information about trial progress or whether proposed CT modifications are likely to improve the chance of success. ROLES OF REGULATORY PROFESSIONALS The RP has many responsibilities when involved with interim analysis and DMCs. First, the RP must contribute to the decision as to whether or not a DMC is warranted (see Table 1). If it is warranted, the RP needs to understand the DMC charter and understand how the sponsor s company intends to communicate with the DMC and any other independent entities involved in the interim analysis. The RP is also responsible for review and input into the CT protocol and the SAP before they are finalized. Next, the RP must weigh the pros and cons of seeking regulatory advice on the trial design and SAP, and possibly even the DMC charter, before trial initiation. On one hand, regulatory feedback can be invaluable in greatly increasing the likelihood that the final SAP and trial design will be acceptable when drug approval is finally sought. On the down side, there are no guarantees as to what level of feedback will be provided nor how long it will be before the regulatory feedback is provided. For example, special protocol assessments by the FDA, particularly of phase 3 trials, might deliver valuable input on trial design, but does the delay to the development timeline by awaiting this feedback outweigh the benefit? If regulatory feedback is sought, the RP must decide how different regulatory agencies (FDA, EMA, etc) need to provide scientific input to ultimately support a global marketing plan. The RP should follow established procedures that outline how the RP will interact with the steering committee, the sponsor s product team, and the DMC both during and following the interim analysis process. Finally, the RP will be responsible for establishing the processes by which regulatory agencies will be informed of interim and final trial results. I N T E R I M A N A LY S I S P I T F A L L S T O A V O I D POOR PRIOR PLANNING One of the many responsibilities of the regulatory professional is to ensure that the role of the DMC and the conduct of any interim analysis is thoroughly and carefully planned out prior to any such analysis. Monitoring boundaries for overwhelming benefit or demonstrated futility must be specified to the DMC in advance, and are at the ultimate discretion of the sponsor with input from the DMC as well as the FDA or other regulatory agencies (10). If the monitoring plan is set up before study start, the DMC should be charged with recommending early stopping on the basis of positive results only when the data are truly compelling and type I error risk is acceptably low. The example of sorafenib, a multikinase inhibitor of tumor cell proliferation and angiogenesis, illustrates how access to interim data may inadvertently damage the scientific integrity of a clinical trial. Sorafenib was approved for the treatment of advanced renal cell carcinoma in December 2005 in an accelerated approval setting. Approval was based on a strong biomarker effect in the interim analysis of the phase 3 trial (n = 903 patients) with sorafenib delaying the increase of disease burden (12). Following release of interim data, control patients were allowed to cross over to the sorafenib arm, despite the fact that the biomarker did not provide a definitive assessment (or correlation) to the primary clinical endpoint, namely progression-free survival. Albeit sorafenib is currently (as of March 2010) approved for the treatment of advanced renal cell carcinoma and resectable hepatocellular carcinoma, final results from the initial confirmatory trial failed to provide interpretable information on overall survival (13) and thus demonstrated the harm to trial integrity that can occur on release of in- Drug Information Journal 06 DIJ 44(6) indd 707 9/7/2010 4:25:52 PM

6 708 r e g u l a t o r y a f f a i r s Coutant terim data. In this example, not crossing over patients following interim analysis, or basing interim looks on survival data, might have saved the integrity of final trial data. Finally, the sponsor is also encouraged to be forthright with the DMC at the time of trial initiation and provide the DMC a thorough recounting of adverse events that have been observed in prior clinical trials. Note that if the FDA were to learn of serious or recurring infractions of this type, the FDA would be entitled to communicate their concerns directly with DMCs during open sessions, or require the sponsor to share the DMC charter with them at time of study initiation. ILL-TIMED INTERIM DATA RELEASE Disclosure of suggestive interim analysis data could critically disturb the equipoise of the investigators. This can have weighty consequences for the conduct of the trial, perhaps leading to a shift in patient randomization by the trial investigators, or negatively affecting adherence to the study protocol in some other manner. Thus, the sponsor should release interim data neither too early nor too late. Do not be misled by short-term benefits and early interim data. The 15,290 patient VALUE trial compared the angiotensin receptor blocker valsartan with a calcium channel blocker, amlodipine, in hypertensive patients at high cardiovascular risk. Interim data suggested the amlodipine control group had a lower incidence of adverse events and risk. The DMC, weighing all the data, did not stop the trial and instead recommended gathering additional data (13). Several years later, final data showed that the short-term risks were not consistent with longer-term relative risks and that it was actually valsartan that led to a substantially lower rate of diabetes. In this case, the DMC did the right thing and preserved CT integrity by not letting short-term data overshadow the important longterm goals. Communicate negative findings in a timely manner. The sponsors failed to present negative findings in a timely manner following final analysis of the ENHANCE clinical trial, which studied Vytorin (an ezetimibe/simvastatin combination) versus simvastatin alone, according to the plaintiffs in the ongoing civil suit (14). Although this is an example of not releasing final data in a timely manner, the same concerns apply for the timing of the release of interim data. The civil plaintiffs claim that the sponsors (Merck and Co/Schering-Plough) received a great deal of revenue from Vytorin sales during the approximately 23-month period between the conclusion of the analysis and the final disclosure of the data. The plaintiffs also cited concerns with the sponsor s plans to change EN- HANCE s primary endpoint postanalysis. The Vytorin class action suit (14) claims that it was only after a congressional letter of concern was sent to the sponsors that the data, based on the original endpoints, was made publicly available. TYPE I ERROR When multiple interim looks at the data are expected, proper statistical methods (such as group sequential methods) are needed to protect against adverse impact of type I error, and the study protocol should outline these methods (3). Consider that for a P value <0.05, it means that an observed effect, or a result even more extreme, has less than 5 chances out of 100 of having occurred if the null hypothesis of no real effect were true. However, following multiple looks at the data, the chance of a false positive result will be considerably higher than 5 out of 100. In other words, experimenters who repeatedly test their accumulating data for a given level of nominal significance will, if they go on long enough, eventually achieve it (15). The sponsor is urged to remember that formal statistical techniques for adjusting P values are only valid if interim analysis is planned a priori. Retrospective P-value adjustments are purely arbitrary for unplanned interim analyses. It is therefore strongly encouraged that any plans for interim analysis are spelled out in trial protocols, including who will (and who will not) have access to the interim data. Note that adjustment of P 06 DIJ 44(6) indd 708 9/7/2010 4:25:52 PM

7 Interim Analysis r e g u l a t o r y a f f a i r s 709 values is not needed if there is no interim analysis option for early trial termination (16). M A K I N G D E C I S I O N S B A S E D O N I N T E R I M A N A LY S I S EARLY STOPPING Ethically, trials must be stopped for futility if study methodology cannot be sustained or if it becomes evident that study objectives cannot be achieved. Trials may also be stopped early if the efficacy signal is great enough that it would be unethical to continue to administer patients the inferior comparator drug (eg, for unmet medical need in the treatment of life-saving or life-sustaining medicines) When considering stopping for efficacy, it is very important that the sponsor sets a predefined statistical stopping boundary for the primary outcome. This boundary should define a strong treatment difference in primary outcomes with a very small P value to avoid type 1 error. However, statistical evidence should not be the sole determinant for early trial termination. It is suggested that harm as well as efficacy be examined by the DMC before making a recommendation for early stopping. Goodman (17) pointed out that if the nature of the harm is known before the study but its frequency is uncertain, then including harm as part of the SAP, and monitoring for the harm by the DMC, may be warranted. Other considerations for early stopping include: Stop early, and potential long-term benefits of a chronic-use drug may be missed. Stopping too early for short-term benefits may not be the most prudent course remember in the end that label language needs to be backed with substantial evidence of both safety and efficacy. For most drugs, this means that two adequate and well-controlled trials (AWCTs) are required to gain desired label language about efficacy and safety in a given population. There is a small number of situations (unmet need and lifesaving drugs) where substantial evidence can be established based on a single AWCT (18), but the majority of drugs need more than a single clinical trial stopped early for clinical benefit. Stop early, and you will never have the longer-term data from that study. If the trial continued, would the effect persist, strengthen, or even reverse? Stopping early almost always leads to a loss of information, including loss of valuable subgroup information (19). Stopping early? Consult the FDA first. Ideally, the sponsor should begin an early dialogue with FDA, at least as early as the time of initial study design, when the statistical monitoring plan and early stopping boundaries are being developed (1). Then, if the DMC recommends early stopping during the trial, it is recommended that the FDA be consulted again. The FDA will want to know if the interim safety data is adequate, if subgroups have been sufficiently examined, and if long-term benefits and secondary endpoints have been adequately considered (20). The discussion of early stopping is concluded with two case examples one good, and one not so good. The first example is the early stopping of the large phase 3 torcetrapib/atorvastatin combination trial in This clinical trial was set up correctly from the start, with an independent DMC and clearly delineated stopping rules. The DMC routinely monitored a number of outcomes (including stroke, heart attacks, and revascularizations) to ensure patient safety. On the morning of December 2, 2006, the DMC notified the sponsor, Pfizer, of an increased rate of mortality observed in the combination treatment compared to atorvastatin alone. By the end of the same day, the FDA was notified of Pfizer s intent to stop the trial, and the FDA fully supported the sponsor s decision to stop the trial. The VIGOR trial, and the DMC monitoring thereof, provides an example of how safety monitoring could be better handled (21). The VIG- OR trial was a large multicenter Vioxx trial that was completed in early In December 1999, three and a half months before the VIG- OR trial ended (and 5 years before Vioxx was pulled from the market for cardiovascular risk), interim data available to the DMC indicated that Vioxx patients suffered twice as many heart attacks, strokes, and deaths as patients on the Drug Information Journal 06 DIJ 44(6) indd 709 9/7/2010 4:25:52 PM

8 710 r e g u l a t o r y a f f a i r s Coutant naproxen active control arm. When asked years later why the trial was not stopped early, the DMC indicated that at the time, they could not be sure if the differences were due to some adverse effect of Vioxx or due to the lack of protective cardiovascular properties of naproxen in the Vioxx arm (21). However, critics contend that naproxen has never been clinically proven to prevent heart attacks and that, in any case, the study should have been stopped if one arm was at double the risk of the other. Another DMC-cited reason was that a low number of heart problems was observed, and that diagnoses from patients and their primary doctors could be wrong. DMC critics said that although numbers were small, similar and consistent trends were observed across several subpopulations of VIGOR, suggesting that the risks were real and that a proper adjudication committee would have helped ensure against diagnosis error. Finally, the DMC mentioned in their defense that the study was nearing completion anyway. Critics have ceded this point, but asked in return where the ethical line for early stopping should have been drawn. ALTERING STUDY DESIGN With regard to the goal of ultimately achieving regulatory approval, there are two types of changes to study protocols that can be made following an interim look at unblinded data (whether by the sponsor or by a DMC). Safe or simple changes would include administrative changes to increase power, feasibility, or compliance (eg, extend recruitment, change CRO central lab, etc). Complex or risky changes would include such things as changes in primary endpoint or sample size and must be carefully reasoned before being implemented. For the latter type of change, the statistician making the recommendation must be blinded to the study and implement statistical corrections if the change is likely to increase type I error. There is precedent that if a sponsor statistician has access to unblinded interim data and yet also sits on the steering committee, the FDA will view the mere presence of the statistician as increasing type I error and multiplicity, and this is exemplified by cautionary language in Section 6.4 of the 2006 FDA Guidance on DMCs (1), which urges sponsors not to allow such a scenario to develop. Complex changes, such as to primary endpoint, should be discussed with the FDA prior to implementation as there is a likely resulting chance that so doing could impact the validity of the trial or support the desired regulatory decision (1). For this reason, regulatory professionals should be included in the decision-making process during protocol change or SAP change deliberations, and to amend the IND or clinical trial agreement if warranted. Despite the difficulties in making complex changes to the ongoing trial following review of interim data, not so doing can also be foolhardy. Failure to make reasonable changes in light of unanticipated events can waste information (22). For instance, the 1989 Physician s Health Study (a trial of 22,000 physicians) compared aspirin against placebo. The study looked for reduction in total mortality. The DMC recommended early stopping because the event rate was much lower than expected. At time of stopping, myocardial infarction (MI) rate showed a dramatic difference. However, in spite of this large and highly significant reduction, the FDA did not approve the MI indication because MI was not the prespecified endpoint. Note that adaptive-design CTs proactively anticipate changes to be made to the study protocol and trial conduct, based either on blinded data or prospectively planned analyses of interim unblinded data. Before the start of an adaptive-design CT, many questions must be answered, including the number of interim analyses required, sample size, and how to design the interim analyses in such a manner that type I error is appropriately controlled (23). However, revisions based on blinded interim evaluations of data (eg, aggregate event rates, variance, discontinuation rates, baseline characteristics, etc) do not usually introduce type I error. At the end of the day, the trial needs to be interpretable at its conclusion and the operational bias needs to be minimized. As stated by Bob Temple (FDA) in 2006, We [the FDA] have no objection to a variety of adaptive randomiza- 06 DIJ 44(6) indd 710 9/7/2010 4:25:53 PM

9 Interim Analysis r e g u l a t o r y a f f a i r s 711 tion designs... but such designs are very infrequently proposed (19). ENRICHMENT: INCREASING SIZE OF STUDY Sponsors plan trials so that they are appropriately powered (eg, sufficient sample size) to provide a reasonable chance of success based on data at hand. In some cases, because of an underestimation of either treatment effectiveness or variability in patient response to treatment, proof of effectiveness cannot be established unless trial size is increased. While it is possible for an independent DMC to be the group that recommends an increase in sample size, this is not ideal, as knowledge of treatment effect can place the DMC in an uncomfortable position (22). Another possibility is to design the study to be event based, rather than based on a set number of patients. Granted, this increases the complexity of the trial because the patient numbers needed are no longer fixed. However, in this case, the sponsor (or DMC) can look at the number of events without unblinding data, and easily increase sample size if merited. Another possibility would be to conduct an interim analysis in a randomized long-term trial looking for patient data with regard to an early response biomarker (eg, tumor shrinkage in oncology trials), and then enrich the trial population with those that showed an early response (19). C O N C L U S I O N Proper planning and conduct of interim analyses are critical to the success of many phase 2 and 3 clinical trials, but particularly so for pivotal trials with DMC involvement. In these cases, DMCs review the unblinded clinical safety and efficacy data. In conjunction, the regulatory professional contributes to the design and implementation of the interim analysis plan that will be utilized by the DMC. The regulatory professional is cautioned that during interim reviews, DMCs need to examine all pertinent efficacy and safety data to fully understand the benefit/risk ratio. DMCs can make safe administrative suggestions to improve recruitment or compliance with the study protocol to aid the study without placing trial integrity in jeopardy. If more substantive changes are made (eg, changes to primary endpoint or sample size), or early stopping of the trial is considered, it is strongly recommended that the FDA, or relevant regulatory body, be consulted prior to implementation. Preferably, boundary criteria for early stopping should be dependent only on major endpoints (eg, mortality) or on surrogate markers with proven strong correlation to the endpoint of interest. When considering early trial termination, remember to consider all ramifications of not collecting longer-term data from that trial. It may be impractical or impossible to repeat the study. For instance, if a planned 5-year trial is stopped after 2 years, obviously longer-term 5-year safety and efficacy data will not be collected; there are many precedents for a drug s benefit/risk ratio to substantially change when longer-term safety and efficacy data become available. Finally, remember that proof of substantial evidence of safety and efficacy (typically more than one adequate and well-controlled trial) is an FDA requirement for approval of any drug or biologic. Any pharmaceutical seeking first approval, or approval of a new indication, must still meet the substantial evidence expectation even if the pivotal trial was unblinded or stopped early. Acknowledgments The author would like to thank Dr. Gregory Enas, Senior Director, Global Regulatory Affairs, Eli Lilly and Company, for his comments and assistance, and also thank the anonymous reviewers whose comments improved the manuscript. R E F E R E N C E S 1. US Food and Drug Administration. Guidance for industry: guidance for clinical trial sponsors: establishment and operation of clinical trial data monitoring committees. March plianceregulatoryinformation/guidances/ ucm pdf (accessed November 9, 2009). 2. World Health Organization. Guidance: operational guidelines for the establishment and functioning of data and safety monitoring boards (ver- Drug Information Journal 06 DIJ 44(6) indd 711 9/7/2010 4:25:53 PM

10 712 r e g u l a t o r y a f f a i r s Coutant Q1 sion 05.1) publications/publications/pdf/operat_guide lines.pdf (accessed November 6, 2009). 3. European Medicines Agency, Committee for Medicinal Products for Human Use. Guidance: guideline on data monitoring committees. July 27, EMEA/CHMP/EWP/5872/ en.pdf (accessed November 9, 2009). 4. National Institutes of Health. NIH policy: policy for data and safety monitoring. June 10, not html (accessed November 9, 2009). 5. International Conference on Harmonization. ICH guideline for good clinical practice, E6 (R1). May html (accessed March 31, 2010.) 6. Temple R. What trials? January 17, 2003, presentation at Issues in Data Monitoring Committees Meeting. 7. Delgado-Herrera L, Anbar D. A model for the interim analysis process: a case study. Control Clin Trials. 2003;24: Ellenberg S. Clinical trial data monitoring committees: consensus and debate. Presentation at the International Conference on Drug Development, February 25, DeMets DL, Fleming TR. The independent statistician for data monitoring committees. Stat Med. 2004;23: Data monitoring committees in practice. Appl Clin Trials Online. March 1, clinicaltrialsonline.findpharma.com/applied clinicaltrials/feature+article/data-monitoring -Committees-in-Practice/ArticleLong/Article/ detail/87247?ref=25 (accessed November 9, 2009). 11. Fisher MR, Roecker EB, DeMets DL. The role of an independent statistical analysis center in the industry-modified National Institutes of Health model. Drug Inf J. 2001;35: europe.diahome.org/diahome/resources/con tent.aspx?type=eopdf&file=%2fproductfiles%2f 8357%2fdiaj_11700%2Epdf (accessed November 9, 2009). 12. Escudier B, Eisen T, Stadler WM, et al. Sorafenib in advanced clear-cell renal-cell carcinoma. N Engl J Med, 2007;356: Fleming TR, Sharples K, McCall J, Moore A, Rodgers A, Stewart R. Maintaining confidentiality of interim data to enhance trial integrity and credibility. Clin Trials. 2008;5: Vytorin-Class-Action website. January 17, (accessed December 7, 2009). 15. McPherson K. Statistics: the problem of examining accumulating data more than once. N Engl J Med. 1974;290: PMA Biostatistics and Medical Ad Hoc Committee on Interim Analysis. Interim analysis in the pharmaceutical industry. Control Clin Trials. 1993;14(2): Goodman SN. Stopping at nothing? Some dilemmas of data monitoring in clinical trials [comment]. Ann Intern Med. 2007;146: O Neill R. Some FDA perspectives on data monitoring in clinical trials in drug development. Stat Med. 1991;12: Temple R. FDA perspective on trials with interim efficacy evaluations. Stat Med. 2006;25: Weschler J. FDA promotes DMCs to enhance subject safety. Appl Clin Trials Online. February 1, pharma.com/appliedclinicaltrials/view+ from+washington/fda-promotes-dmcs-to -Enhance-Subject-Safety/ArticleStandard/Arti cle/detail/87246 (accessed November 9, 2009). 21. Prakash S. Conflicted safety panel let Vioxx study continue. National Public Radio transcript (June 8, 2006, broadcast). 22. Wittes J. On changing a long-term clinical trial midstream. Stat Med. 2007;27: O Neill R. Adaptive designs: musings and myths. Opportunities, challenges and scope in drug development. November 13 14, _Designs_Presentations/09_Robert_Oneill _Myth_Busting_Statistical.pdf (accessed November 20, 2009). The author reports no relevant relationships to disclose. \Q1\Please identify location, date, and sponsoring organization for the meeting. 06 DIJ 44(6) indd 712 9/7/2010 4:25:53 PM