Stage-wise adaptive designs

Free download. Book file PDF easily for everyone and every device. You can download and read online Stage-wise adaptive designs file PDF Book only if you are registered here. And also you can download or read online all Book PDF file that related with Stage-wise adaptive designs book. Happy reading Stage-wise adaptive designs Bookeveryone. Download file Free Book PDF Stage-wise adaptive designs at Complete PDF Library. This Book have some digital formats such us :paperbook, ebook, kindle, epub, fb2 and another formats. Here is The CompletePDF Book Library. It's free to register here to get Book file PDF Stage-wise adaptive designs Pocket Guide.

Additionally, non-inferiority to placebo with respect to the need for the rescue thiopentone use has to be shown hypothesis 3.

Why, what and when to adapt in clinical trials

It was a study of the PAED-Net, which is a corporation of pediatric modules in six German university locations in a small and vulnerable population. Regulatory authorities usually expect statistical significance in all three co-primary endpoints simultaneously, which would mean that no further multiple testing correction is needed. But in the setting of pediatric populations the investigators were convinced that under double blind conditions it would be worth to achieve significance in at least one of the two superiority hypotheses.

The corresponding global null hypothesis was tested using the OLS test, and a closed testing was planned in order to show significant differences in specific endpoints []. The study was planned in three stages with critical values according to O'Brien and Fleming adjusted significance levels 0. The results of the three stages were combined using the inverse normal method together with the closed testing principle as described in Section 3. Overall power was defined for detecting at least one significant difference. Considering the power to reject all hypotheses required a lot more patients and was considered inappropriate.

Shop now and earn 2 points per $1

The last option seemed to be useful because it was not clear at the beginning if both fentanyl and midazolam consumption could be reduced with clonidine. The first interim analysis yielded very promising results: The overall p -value for the OLS test was 0. This trend was dramatically confirmed at the second interim analysis: A negative effect in midazolam was observed for the second stage data, yielding ap -value of 0. Nevertheless, the OLS test for the global multivariate hypothesis yielded a p -value of 0. This was statistically significant, although a study continuation was recommended because superiority with regard to fentanyl was not very clear anymore.

Particularly, it was not significant within the closed test procedure, only noninferiority with regard to thiopentone was significant within the closed test procedure. The study recommendation was also to drop midazolam consumption as a primary endpoint from the further analysis because it might jeopardize an overall positive result of the study. Although a positive result was likely for the reduced clinical question, at the final analysis, even fentanyl could not be shown to be significant.

The main study results were published in []. The example shows the importance of keeping adaptive interim decisions secretly. If the doctors had been aware of the fact that midazolam was dropped from further confirmatory analyses, this might have clearly influenced treatment and medication of the patients.

In order to exclude this possibility, only the Independent Statistical Center and the IDMC who gave the recommendation were aware of the study results. The head of the study was informed that there was a recommendation, the decision on it was left to one representative of the sponsor Boehringer Ingelheim, Germany that needed to be involved. This again illustrates that trial logistics is an important issue in adaptive designs. It needs to be extensively discussed how operational biases can occur and how they can be avoided. Although the study did not show the desired effect, especially the second interim analysis illustrates the potential advantage of an adaptive way of analyzing data.

There were no strict stopping criteria, and the continuation of the trial produced a disappointing result but obviously reflects reality. The study was planned in and finalized in Publication of the non-convincing study results was a problematic issue. Another issue with publishing complicated adaptive designs in medical journals is to have space to communicate the statistical methodology []. Several details of the statistical study design were not provided in [].

We nevertheless think that this study serves as an interesting example for an early attempt for an adaptation, which goes beyond sample size reassessment and treatment arm selection in a vulnerable, small population []. Nowadays, the availability of software is a necessary condition for the applicability and acceptance of a statistical methodology.

Many of the procedures proposed for adaptive designs additionally require high-computational effort such that software should be able to perform time consuming computations in a relatively short time. Up to now, the reviews of software packages on clinical trials with interim analysis concentrated on packages specifically designed for group sequential methods [], the reason simply being that software for adaptive designs was not available at that time.

One review of software for adaptive designs [] appeared recently. There is wide field of available commercial software for group sequential designs e. Christopher Jennison provides Fortran programs for group sequential designs freely available on his homepage: www. Further, Fortran programs for the computation of the use function approach are available from the University of Wisconsin School of Medicine and Public Health site www. Specifically for confirmatory adaptive designs, there is still only a limited number of available software, both commercially and non-commercially.

It is commercially available since as a stand alone tool for designing, simulating and performing analysis with an emphasis on the adaptive confirmatory technique. The MC module provides additional multiple comparison features for more than two treatment arms in simulation and analysis, and the PE module additional features for patient enrichment designs in simulation and analysis.

There is also the new DF module with capabilities for adaptive dose-finding designs. East from Cytel www. We list the packages with a short description. It is emphasized that this is a dynamic development, and we expect many more packages in the near future. The functions defined in this program serve for implementing adaptive two-stage adaptive tests that are based on the combination testing principle. This module enables the calculation of confidence intervals in adaptive group sequential trials. This package runs simulations for adaptive seamless designs with and without early outcomes for treatment selection and population enrichment type designs.

This is an interactive tool for designing and evaluating certain types of adaptive enrichment designs. Furthermore, the book [] contains some R-programs for adaptive designs. It also includes programs for performing sample size reassessment procedures and some basic adaptive randomization designs. The book also comes with SAS macros, the most of them performing simulations for the adaptive designs described in the book. To summarize, some software is free and hence attractive for statistical research. This is particularly true for the increasing number of available R-packages.

Simulation-based evaluation of operating characteristics of adaptive designs is becoming increasingly important, some of the available adaptive R-packages already include such functionality. Adaptive confirmatory designs have shaken the classical design paradigm that the details of the design and statistical analysis all have to be all laid down in advance. There are two ways on how to achieve the flexibility of mid-trial design modifications, which in principle need not to pre-specified.

In the first approach, separate test statistics such asp -values or z-values are calculated from the disjoint samples at the different stages and combined into an overall test statistics in a fixed pre-defined way. This approach relies on invariance properties of the stagewise test statistics, for example, that the p -values under the null hypothesis are uniformly distributed. In the second approach, the probability of an erroneous rejection of the null hypothesis by the pre-planned design given the data at interim conditional error rate is calculated at the interim analysis.

The remainder of the pre-planned design then can be replaced by any other design, which never would raise a larger conditional error rate.

  • Cation Binding by Humic Substances (Cambridge Environmental Chemistry Series)?
  • The Thief (Isaac Bell, Book 5).
  • Restructuring Science Education: The Importance of Theories and Their Development!
  • Information theory, evolution, and the origin of life.
  • The Life of Martin Luther King, Jr.. Leader for Civil Rights.

Both methods allow a frightening multitude of flexibility, some of it having been sketched in this review. However, in case adaptations are made by leaving the setting of group sequential designs, both approaches use test statistics that are different from the minimal sufficient test statistics of conventional tests. Curious examples can be constructed that. One way to avoid such scenarios is to plan for some reasonable constraints on flexibility or to allow a rejection of the null hypothesis only if also the conventional test statistics would reject, for example, at full level a.

Clearly, this latter 'dual test' comes at the cost of a slight decrease in power but avoids borrowing strength from a curious weighting of the stagewise test statistics [55,56,,]. Whenever the conditional error of the adapted design using the conventional analysis is lower than the one of the originally planned design, the dual test is implicitly applied.

This property is used, for example, by Mehta et al. Note that the well-founded sufficiency principle might obviously be considered as inappropriate or at least too strongly formulated in the context of providing a statistical methodology for clinical research. The choice of a methodology is not only guided by mathematical optimality but also by practical demands, and the price for adaptivity is the concession to use a suboptimal test statistic in case an adaptation was performed.

Interestingly, the Bayesian principle of combining prior beliefs with new evidence from data and thereby allowing adaptivity by definition might be better suited for the practical demands but lacks type I error rate control, at least in general. Nevertheless, Bayesian methods [73] can be used for assessing interim results e. If hypotheses are adaptively changed during an ongoing trial, multiple testing adjustments will be needed in order to get valid inference on the hypotheses under investigation, allowing conclusive interpretation of the study results.

This is a basic requirement of scientific rigor not to bypass statistical principles by abusing flexibility in a wrong direction. It is not surprising that multiple testing in connection with adaptive choice of hypotheses introduce a further level of methodological complexity. Also, estimation is not a simple problem in adaptive design. A lot of research over the last years have tackled these issues and has given us a better understanding of inference in flexible designs.

But obviously, there is another challenge that arises from how to apply adaptive designs in the environment of clinical trials. It took quite a time till group sequential designs have found their way into clinical trial routine. Nowadays, interim analyses, at least to decide whether to stop for futility, have become an important feature of large clinical trials to account for ethics and costs.

Clearly, the logistics and workload for unblinded adaptive interim analyses who is doing it, what information is passed on, who is getting informed and who is deciding on adaptations? Hence, the required input may be prohibitive to consider adaptive designs as an option in several standard clinical trial scenarios.

The adaptive approach was essentially laid down by the seminal papers in the 90's of the last century. It was accomplished by some kind of natural skepticism from the methodological and the regulatory perspective. In the meantime, there is some clarification of how and when to use adaptive designs. Some concern is still being caused by the unblinding of study results at interim stages. This is an essential feature of classical group sequential designs, and hence part of criticism on adaptive designs is inherited from the latter approach. However, adaptation must not necessarily be based on breaking the blind.

One of the simplest adaptive designs is the blinded SSR design that usually consists of two stages in which the sample size for the second stage is determined based on the first-stage data. This was introduced as the 'internal pilot design' at more or less the same time [8,9]. The blinded SSR design determines the sample size of the second stage using only the estimate of nuisance parameters such as variance or overall standard deviation, overall response rate, or overall survival pattern see, e.

This design is easy to implement and generally does not require adjustments. It is quite efficient if the true treatment effect is close to the pre-planned target that is fixed in such designs. However, the pre-planned target treatment effect is often based on an optimistic guess of the 'clinically relevant' treatment difference at which the power is specified. We think that even for the unblinded cases, a careful consideration of both the effect size and the variability should serve as a guideline for making interim decisions including SSR.

There is no such thing as a free lunch. Therefore, in planning a study, it should be carefully checked in advance, whether in the specific situation the existing caveats are dominated by the potential benefit of flexibility. It has to be kept in mind, however, that in a specific scenario, a method allowing flexibility of going beyond the specific scenario will not be able to beat a method that is optimally tailored to just that one scenario.

The advantage of the group sequential version of adaptive designs is that if no adaptations are made at interim, the 'optimally' planned design will be performed without paying any price for the option of flexibility. The price would be paid if the adaptation detracts the experimenter from the truly optimal design. Note, however, that all these different situations are rather theoretical and can hardly be. The fundamental pros and cons have been exchanged.

Some early enthusiasm from clinical trialists who hoped to get new useful tools may also have been caused by not fully understanding the inferential complications created by flexibility. This may also have been a motivation of regulators in their guidance documents to set rather narrow limits for applying unblinded adaptive designs in drug development e.

Hence, as for many other research dealing with innovative treatments of patients naive promises arising at the beginning here and there did not become real in the present clinical trial community. There are successful examples of carefully planning and running adaptive clinical trials, which the proponents considered to have been helpful in drug development.

So the proof of principle on how to use the methodology for different types of adaptations has been given. By the way, there are also other areas like genomics with huge number of hypotheses where some basic ideas of the adaptive testing principle have been taken over. It should also be kept in mind that adaptive designs are a very potent tool to deal with the unexpected in clinical trials. Not all the details necessary for optimally designing a clinical trial are available in the planning phase. If they were available rather, no trial would be planned at all.

Whenever serious deviations from the planning assumptions become obvious at interim, it may be advantageous as an 'ultima ratio' to use the adaptive design methodology for overcoming deficiencies and 'saving' the running trial, for example, by applying the conditional error rate approach.

Online Program

This would be a reasonable way to deal with protocol amendments containing substantial design modifications in a scientifically honest way. Protocol amendments in practice are an extensively used tool often to open a short cut for introducing flexibility in ongoing clinical trials, which is not accounted for in the statistical analysis. Beyond all skepticism, there are opportunities where adaptive design serve as a valuable tool and generalization of commonly used methods.

In many of the cases where unblinded interim analyses are performed, there is the potential to extend these to include data driven changes of the design. Clearly they cannot be applied and even do not help for all cases. For example, maintaining type I error rate control using all interim information might be critical. In cases where patients for further stages were already recruited and data which is related to the endpoint is already available at an earlier stage, type I error rate control is violated.

This is specifically the case in survival designs. As briefly reviewed here, providing solutions is the topic of current research. Although there are methodological caveats, the incorporation of adaptive elements in study designs seems to work in general. However, sometimes, it is even not preferable to incorporate interim results, especially from small stages, for example, because information from interim results might be insecure, and it might be better to stick to the original assumptions rather than trusting the interim results [57,], or a treatment arm or population is too often wrongly selected and therefore yields power smaller than a conventional design.

This is not surprising but often overseen when people want to look at the data as soon as possible. Designing an adaptive design is a complex task and more statistical planning is needed to account for the flexibility introduced in the study design. Specifically, statistical software is increasingly needed to evaluate the adaptations and to find reasonable strategies.

The question might arise if potential decisions made at interim stages might not be better placed to the upfront planning stage. Adaptive designs also come along with more operational and organizational problems. If the sponsor is not able or not willing to recruit an additional number of patients, it makes no sense to ask for sample size increase. In this case, stopping for futility might be the only option that is worthwhile to consider. If the sponsor agrees to a potential sample size increase, the randomization process including drug supply needs to be managed. Is it possible to correctly administer the drug according to a selected dose regimen?

How does the concrete adaptation work? This makes the dissemination of study results an issue and the careful formulation of the information flow, for example, in an IDMC charter, becomes an essential part of designing the trial. Some feasible models for implementation of adaptive designs have been proposed by the industry []. In summary, adaptive designs have been carefully developed in the past 25 years and - at least from a theoretical perspective - their properties, advantages and disadvantages, are well understood. To achieve full acceptance in the statistical community and by regulators, there is still the need for both more methodological expertise [] and practical experience.

Is Adaptive Design Right for You?

The more this exciting methodology will be used the more it will be understood when it is helpful and when it is not. We thank the editor Ralph D'Agostino for the invitation to write this review. This project has received funding from the European Union's Seventh Framework Programme for research, technological development, and demonstration under grant agreement no IDEAL - Integrated Design and Analysis of small population group trials. Bauer P. Multistage testing with adaptive designs.

Biometrie und Informatik in Medizin und Biologie ; Evaluation of experiments with adaptive interim analyses. Biometrics ; , correction in Biometrics , Adaptive designs in clinical drug development -an executive summary of the PhRMA working group. Journal of Biopharmaceutical Statistics ; Bauer P, Kieser M. Combining different phases in the development of medical treatments within a single trial. Statistics in Medicine ; Adaptive clinical trial designs for European marketing authorization: a survey of scientific advice letters from the European Medicines Agency.

Trials ; 15 1 Food and Drug Administration. Wittes J, Brittain E. The role of internal pilot studies in increasing the efficiency of clinical trials. Internal pilot studies for estimating sample size. Friede T, Kieser M. Sample size recalculation in internal pilot study designs: a review. Biometrical Journal ; 48 4 Posch M, Proschan MA. Unplanned adaptations before breaking the blind. Statistics in Medicine ; 31 30 : Blinded sample size re-estimation for recurrent event data with time trends.

Statistics in Medicine ; 32 30 Connections between permutation and t-tests: relevance to adaptive methods. Statistics in Medicine ; 33 27 Tsiatis AA, Mehta C. On the inefficiency of the adaptive design for monitoring clinical trials. Biometrika ; Jennison C, Turnbull BW.

Medidata Blog

Mid-course sample size modification in clinical trial. Efficient group sequential designs when there are several effect sizes under consideration. Statistics in Medicine ; 25 6 Adaptive and nonadaptive group sequential tests. On the efficiency of adaptive designs for flexible interim decisions in clinical trials. Journal of Statistical Planning and Inference ; 6 Lehmacher W, Wassmer G. Adaptive sample size calculations in group sequential trials.

Biometrics ; 55 4 : Modification of sample size in group sequential clinical trials. Biometrics ; Designed extension of studies based on conditional power. Adaptive group sequential designs for clinical trials: Combining the advantages of adaptive and of classical group sequential approaches. A general statistical principle for changing a design any time during the course of a trial.

Bayesian Adaptive Methods for Clinical Trials. HuF, Rosenberger WF. Optimal restricted two-stage designs. Controlled Clinical Trials ; Internally adaptive designs for parallel group trials. Drug Information Journal ; Stallard N, Todd S. Sequential designs for phase III clinical trials incorporating treatment selection. Statistics in Medicine ; 22 5 Maximum type 1 error rate inflation in multiarmed clinical trials with adaptive interim sample size modifications.

Biometrical Journal ; 56 4 Flexible sample size considerations using information-based interim monitoring. Information-based sample size re-estimation in group sequential design for longitudinal trials. Statistics in Medicine ; 33 22 Type I error rate control in adaptive designs for confirmatory clinical trials with treatment selection at interim.

Pharmaceutical Statistics ; 10 2 Sequential tests of hypotheses in consecutive trials. Biometrical Journal ; Recursive combination tests. Journal of the American Statistical Association ; Hartung J. A self-designing rule for clinical trials with arbitrary variables. Wassmer G. Multistage adaptive test procedures based on Fisher's product criterion. Posch M, Bauer P. Adaptive two stage designs and the conditional error function.

A comparison of two methods for adaptive interim analyses in clinical trials. Brannath W, Bauer P. Optimal conditional error functions for the control of conditional power. Inference on multiple endpoints in clinical trials with adaptive interim analyses. Hommel G. Adaptive modifications of hypotheses after an interim analysis. A unified theory of two-stage adaptive designs.

Liu Q, Pledger GW. On design and inference for two-stage adaptive clinical trials with dependent data. Journal of Statistical Planning and Inference ; Probabilistic foundation of confirmatory adaptive designs. Conditional rejection probabilities of Student's t-test and design adaptation. Increasing the sample size during clinical trials with t-distributed test statistics without inflating the type I error rate. Statistics in Medicine ; 26 12 An approach to the conditional error rate principle with nuisance parameters.

Biometrics ; 67 3 Type I error in sample size re-estimations based on observed treatment difference. Statistics in Medicine ; 20 4 A comparison of methods for adaptive sample size adjustment. Statistics in Medicine ; 20 24 Increasing the sample size when the unblinded interim result is promising. Statistics in Medicine ; 23 7 Denne JS. Sample size recalculation using conditional power. Issues in designing flexible trials.

Statistics in Medicine ; 22 6 The reassessment of trial perspectives from interim data - a critical view. Statistics In Medicine ; On closed testing procedures with special reference to ordered analysis of variance. Tutorial in biostatistics: Adaptive designs for confirmatory clinical trials. Sequential and multiple testing for dose-response analysis. Testing and estimating in flexible group sequential designs with adaptive treatment selection.

Adaptive seamless designs: selection and prospective testing of hypotheses. Journal of Biopharmaceutical Statistics ; 17 6 On sample size determination in multi-armed confirmatory adaptive designs. Journal ofBiopharmaceutical Statistics ; Friede Tim, Stallard Nigel. A comparison of methods for adaptive treatment selection. Biometrical Journal ; 50 5 Stallard N, Friede T. A group-sequential design for clinical trials with treatment selection. Flexible sequential designs for multi-arm clinical trials. Adaptive sequential testing for multiple comparisons.

  • 10 Good Questions About Life And Death;
  • Stage-Wise Adaptive Designs?
  • Forty-Five Etudes for the Oboe?
  • 1938: Hitlers Gamble!

Journal of Biopharmaceutical Statistics ; 24 5 Adaptive Dunnett tests for treatment selection. Confirmatory adaptive designs with Bayesian decision tools for a targeted therapy on oncology. Pharmaceutical Statistics ; Approaches to evaluation of treatment effect in randomized clinical trials with genomic subset. Pharmaceutical Statistics ; 6 3 Adaptive patient enrichment designs in therapeutic trials. Biometrical Journal ; 51 2 Adaptive designs for subpopulation analysis optimizing utility functions.

Biometrical Journal ; 57 1 Wassmer G, Dragalin V. Designing issues in confirmatory adaptive population enrichment trials.

Stage-Wise Adaptive Designs | Jefferson Campus Store

Journal of Biopharmaceutical Statistics ; early view. Optimizing trial design sequential, adaptive, and enrichment strategies. Circulation ; 4 Mehta C, Gao P. Help Centre. My Wishlist Sign In Join. Be the first to write a review. Add to Wishlist. Ships in 7 to 10 business days. Link Either by signing into your account or linking your membership details before your order is placed. Description Table of Contents Product Details Click on the cover image above to read some pages of this book!

Industry Reviews "Covering a broad range of material, this book may serve well as a reference source for adaptive approaches in the design of experiments. Author Index. Subject Index. Keywords multiple testing adaptive designs multiple hypotheses multiple endpoints multiple treatments. Hellmich, Martin; Hommel, Gerhard. Multiple testing in adaptive designs—a review. Abstract Chapter info and citation First page Abstract During the course of a study it would be desirable to take advantage of new internal or external information in order to modify design features laid down in the study protocol.

Chapter information Source Y.

Export citation. Export Cancel. You have access to this content.