Trial Conduct


Because randomization is the signifying characteristic of Phase III trials, correct implementation and maintenance of this feature is vital. In the simplest case, a series of treatment assignments are generated from a random mechanism such as a random number table or computer program. Sequential assignments should only be revealed one at a time (to avoid compromising the randomness with respect to the next assignment), and thus it is preferable to have a secure centralized randomization procedure, via telephone or computer contact. To balance the number of patients in each arm, block randomization, in which an equal number of assignments on each treatment arm of the trial occurs after each block of patients is enrolled, can be used (for example, assigning ABABBA, AABABB, etc., for randomly ordered strings of assignments brings the number on each arm into balance after each set of six patients; note that the size of the block is not revealed and can also be varied to make the process unpredictable). To incorporate stratification factors to be balanced between treatment arms, blocked randomization can be used within each stratum. When the number of strata is large, and in particular in multicenter trials, this type of assignment scheme can become unwieldy and cannot necessarily assure balance. In this case, some type of dynamic allocation scheme can be used whereby the current assignment is generated based on previous assignments. This type of randomization must be centralized, as stratification factor data for all previously enrolled patients must be available when randomizing the current patient. Rather than deterministic assignments to balance the arms, often "biased-coin" randomization is used, in which the assignment probability is weighted toward the arm for which assignments are needed to achieve balance. The minimization algorithm is frequently used for dynamic balancing taking stratification factors into account.58 This allocation scheme balances treatment arms for each stratification factor singly, but not necessarily for every combination of factors, as in fully stratified randomization. However, it is much easier to manage over multicenter studies and performs well for multiple stratification factors.57

There are several ways in which the benefits of randomization can be eroded or nullified. Of course, any breach of the random assignment process has an irreparable effect on the validity of the trial. Second, a large number (or differential number per arm) of patients "canceled" or withdrawn from the trial calls into question the validity of the comparison for the remaining participants. Differential follow-up and ascertainment of status per arm can have a similar effect. Third, bias in assessment of outcomes can have a major impact on the estimated treatment effect, and thus, objective outcome measures and blinding of treatment assignment come into play. Treatment assignment blinding is not feasible for many oncology trials (e.g., radiotherapy and most chemotherapy regimens), but has been used with great success in others (e.g., tamoxifen). In either case, and in particular for studies that cannot be blinded (among patients or caregivers), unambiguous, objectively defined endpoints are essential. In cases where determination of the endpoint involves possible observer subjectivity, such as when reading a diagnostic scan to determine disease progression, keeping assessors unaware of treatment assignment may be necessary.

toward a significant difference in favor of the standard treatment would be continued until such a result was realized. Thus, an additional rule that allows for stopping early for "futility" with respect to the new treatment is usually also specified. Alternatively, asymmetric boundaries that more easily allow stopping early for evidence that the new treatment may actually be inferior to the standard (evidence de facto that the new treatment will not ultimately prevail) can be used. More comprehensive treatment of this topic can be found in clinical trials texts1,2 or books on group sequential monitoring.59 Regardless of the specific approach adopted, an interim analysis plan should be devised during trial design and adhered to throughout trial conduct because failure to account for multiplicity of analyses can result in spurious positive findings. Also, a specified plan may help to avoid diminished influence of a trial even when results are decidedly positive, which can occur if there is a perception by others that the trial had been terminated prematurely due to favorable results at a particular analysis time.60

The decision to discontinue the trial (accrual and/or treatment, depending on its current state) and release findings is typically vested in an independent Data and Safety Monitoring Committee (DSMC). However, in addition to evaluating according to the monitoring rule, the DSMC considers a broader body of information regarding the trial as well as external information that bears on treatment for the disease under study. Table 8.5 outlines the membership, aspects of trial conduct over which the DSMC has oversight, and recommendations that might arise from trial review and DSMC

Trial Monitoring

As in earlier trials, Phase III trials include provisions for formal oversight of risks and benefits to ensure patient welfare and use resources efficiently. Interim monitoring consists of both continuous oversight of adverse events and periodic interim tests (a predetermined number) of primary study hypotheses. With respect to these interim tests, a large body of statistical developments has addressed how to conduct tests to determine if an early determination of treatment superiority is warranted while at the same time controlling for inflation of a error (e.g., false-positive findings) resulting from repeatedly performing statistical hypothesis tests. Caution is also warranted because early results from time-to-event data tend to be unstable and change as more information accumulates. In essence, these problems are addressed by simply requiring a stricter criterion than the typical P less than 0.05 "significance" for interim analyses, and early approaches to this problem used either a smaller constant significance criterion throughout interim and definitive analyses, chosen such that the significance level for the entire set of sequential tests does not exceed a, or a constant but much more stringent criterion early in interim analyses and a more conventional significance level after the trial has accumulated the requisite information for definitive analysis. Figure 8.1 shows these and some other examples of early stopping boundaries, which when exceeded at any of the interim analyses shown on the x-axis would prompt consideration of early stopping. These boundaries are symmetric with respect to superiority for either the new or standard treatment group, reflecting the fact that two-sided hypothesis tests remain the convention. In reality, it is unlikely that a trial tending

Herbal Remedy Secret Uncovered

Herbal Remedy Secret Uncovered

Discover How To Use Herbal Medicine Effectively To Heal Away Disease amp illnesses That Most Of The Herbalist Do Not Want You To Know About. If You Have Never Know What Is All About Herbal Medicines amp The Correct Way Of Using Herbs To Build A Healthier Life, Then This Guide Is About To Reveal All Just That.

Get My Free Ebook

Post a comment