(This is a draft and truncated version - for final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals)
adaptive designs for clinical trials
Adaptive designs allow flexibility to redesign a clinical trial midstream, guided by the interim experience. This can make the trial more efficient by saving time, money and patients. Suppose after due consideration of available knowledge, you plan a clinical trial on 1000 subjects in each arm with one-year follow-up. After two months into the trial, you find from interim analysis that your anticipations at planning stage were incorrect and that the trial will give you confirmatory results on efficacy (or lack of it) in just 6 months (or that you need to extend it to 16 months to get an adequate number with the desired end-point); or only 700 would suffice (or 1500 would be needed); or that doses you are trying are too high (or too low) and you need to have a middling dose; or that a concomitant treatment is needed; or that a particular subgroup such as males of age 70+ years needs to be excluded. Sometimes even the baseline information on the enrolled subjects may indicate that modifications in the design are needed. For example, this may tell you that expected kind of subjects are not being enrolled and eligibility criteria need to be changed. In such situations, you would not like to waste resources by sticking to the original plan in the hope of wonders, instead would like to adapt the trial to the realities revealed by actual experience.
Need for adaptation arises from the fact that sometimes all the information needed to plan a near perfect trial is not available in advance. Beside medical issues such as side-effects, biostatistical issues that can affect planning are the variance of the estimate of the effect-size you wish to consider, expected compliance, and a conjecture of the anticipated effect-size itself. When these are not available with reasonable degree of assurance, you may like to fall back on adaptive design that allows flexibility. Adaptive designs incorporate practical considerations of possibly not getting things right first time when the design was prepared.
Adaptation is planned in advance by anticipating different scenarios that could unfold in the planned trial. Thus, it can handle only limited issues and not those that were unforeseen or ignored. The issues must arise from interim analysis and not from extraneous sources. Changing the features of the trial due to other considerations is not adaptation in the sense we are discussing here. Adaptation is not a substitute for a sloppily planned trial. On the other hand, implicit in all this is that the trial has immaculate design that has considered all the available information, including what adaptations to be made in case of finding specific issues.
Despite clear advantages of adaptive trials, they have been rarely used so far. First difficulty is that the adaptive methodology is still evolving. Secondly, this could involve many features of the design, and can be done possibly at several stages of an ongoing trial. It is difficult to visualize in advance where and what adaptations would be required. Thirdly, there is a confusion of what all should be considered for adaptation, how to keep Type-I and Type-II errors under control despite periodic evaluations, and how to handle logistic problems that arise from adaptation of an ongoing trial. The timing of interim analysis will be stated in the trial’s protocol: it will occur at a predefined time. Care is taken that such appraisal does not undermine the integrity and validity of the trial.
When adaptations are done, the analysis becomes computationally complex, for which specially tailored softwares are coming up. Nonetheless, the adaptive strategy is being explored with enthusiasm by drug companies, regulators and researchers. As the experience accumulates, adaptive trials can be better designed and will have wider acceptability. An adaptive trial can quickly move from phase-II to phase-III since lessons learnt are already incorporated. This can expedite the process of product development. For sure, this strategy can address frustration arising in conventional structured designs when the trial gives negative results and you wish that if something can be done differently the results would not be so disappointing. Adaptive trials may soon become industry standard as the problems in their implementation are sorted out. ...
Englert and Kieser [1] have discussed how adaptive designs in phase-II cancer clinical trials help in reducing the required number of subjects for coming to a decision. For more details of designs with interim appraisals, see Chow and Chang [2].
[1] Englert S, Kieser M. Optimal adaptive two-stage designs for phase II cancer clinical trials. Biom J 2013 Nov;55(6):955-68. http://www.ncbi.nlm.nih.gov/pubmed/23868324
[2] Chow S-C, Chang M. Adaptive Design Methods in Clinical Trials. Chapman & Hall/CRC, 2006.
For final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals
Concise Encyclopedia of Biostatistics for Medical Professionals)
adaptive designs for clinical trials
Adaptive designs allow flexibility to redesign a clinical trial midstream, guided by the interim experience. This can make the trial more efficient by saving time, money and patients. Suppose after due consideration of available knowledge, you plan a clinical trial on 1000 subjects in each arm with one-year follow-up. After two months into the trial, you find from interim analysis that your anticipations at planning stage were incorrect and that the trial will give you confirmatory results on efficacy (or lack of it) in just 6 months (or that you need to extend it to 16 months to get an adequate number with the desired end-point); or only 700 would suffice (or 1500 would be needed); or that doses you are trying are too high (or too low) and you need to have a middling dose; or that a concomitant treatment is needed; or that a particular subgroup such as males of age 70+ years needs to be excluded. Sometimes even the baseline information on the enrolled subjects may indicate that modifications in the design are needed. For example, this may tell you that expected kind of subjects are not being enrolled and eligibility criteria need to be changed. In such situations, you would not like to waste resources by sticking to the original plan in the hope of wonders, instead would like to adapt the trial to the realities revealed by actual experience.
Need for adaptation arises from the fact that sometimes all the information needed to plan a near perfect trial is not available in advance. Beside medical issues such as side-effects, biostatistical issues that can affect planning are the variance of the estimate of the effect-size you wish to consider, expected compliance, and a conjecture of the anticipated effect-size itself. When these are not available with reasonable degree of assurance, you may like to fall back on adaptive design that allows flexibility. Adaptive designs incorporate practical considerations of possibly not getting things right first time when the design was prepared.
Adaptation is planned in advance by anticipating different scenarios that could unfold in the planned trial. Thus, it can handle only limited issues and not those that were unforeseen or ignored. The issues must arise from interim analysis and not from extraneous sources. Changing the features of the trial due to other considerations is not adaptation in the sense we are discussing here. Adaptation is not a substitute for a sloppily planned trial. On the other hand, implicit in all this is that the trial has immaculate design that has considered all the available information, including what adaptations to be made in case of finding specific issues.
Despite clear advantages of adaptive trials, they have been rarely used so far. First difficulty is that the adaptive methodology is still evolving. Secondly, this could involve many features of the design, and can be done possibly at several stages of an ongoing trial. It is difficult to visualize in advance where and what adaptations would be required. Thirdly, there is a confusion of what all should be considered for adaptation, how to keep Type-I and Type-II errors under control despite periodic evaluations, and how to handle logistic problems that arise from adaptation of an ongoing trial. The timing of interim analysis will be stated in the trial’s protocol: it will occur at a predefined time. Care is taken that such appraisal does not undermine the integrity and validity of the trial.
When adaptations are done, the analysis becomes computationally complex, for which specially tailored softwares are coming up. Nonetheless, the adaptive strategy is being explored with enthusiasm by drug companies, regulators and researchers. As the experience accumulates, adaptive trials can be better designed and will have wider acceptability. An adaptive trial can quickly move from phase-II to phase-III since lessons learnt are already incorporated. This can expedite the process of product development. For sure, this strategy can address frustration arising in conventional structured designs when the trial gives negative results and you wish that if something can be done differently the results would not be so disappointing. Adaptive trials may soon become industry standard as the problems in their implementation are sorted out. ...
Englert and Kieser [1] have discussed how adaptive designs in phase-II cancer clinical trials help in reducing the required number of subjects for coming to a decision. For more details of designs with interim appraisals, see Chow and Chang [2].
[1] Englert S, Kieser M. Optimal adaptive two-stage designs for phase II cancer clinical trials. Biom J 2013 Nov;55(6):955-68. http://www.ncbi.nlm.nih.gov/pubmed/23868324
[2] Chow S-C, Chang M. Adaptive Design Methods in Clinical Trials. Chapman & Hall/CRC, 2006.
For final and full version, see
Concise Encyclopedia of Biostatistics for Medical Professionals