Flexible designs of clinical trials – a graphical representation in form of a chance tree

mibe000102 10.3205/mibe000102 urn:nbn:de:0183-mibe0001024 Research Article Flexible designs of clinical trials – a graphical representation in form of a chance tree Flexible Designs für klinische Studien – eine grafische Repräsentation in Form eines Wahrscheinlichkeitsbaums Hellmich Hellmich Martin M PhD

Institute of Medical Statistics, Informatics, and Epidemiology, University of Cologne, Kerpener Str. 62, 50924 Cologne, Germany, Phone: +49 221 478-6509, Fax: +49 221 478-6520Institute of Medical Statistics, Informatics, and Epidemiology, Cologne University Hospital, Cologne, Germany

martin.hellmich@uni-koeln.de author German Medical Science GMS Publishing House

Düsseldorf

610 adaptation chance tree flexible design interim analysis meta analysis Adaptation Flexibles Design Meta-Analyse Wahrscheinlichkeitsrechnung Zwischenauswertung 20100302 engl 1860-9171 6 1 GMS Medizinische Informatik, Biometrie und Epidemiologie GMS Med Inform Biom Epidemiol 02 Hintergrund: Flexible Designs für klinische Studien erlauben eine bessere Anpassung laufender Untersuchungen an aktuelle Umstände, wie z.B. eine unerwartet kleine Effektgröße, ohne die statistische Fehler-Kontrolle zu durchbrechen. Die vorhandenen statistischen Verfahren sind jedoch komplex.Methodik: Ein zweistufiges, flexibles Design wird in Form eines einfachen Wahrscheinlichkeitsbaums repräsentiert. Kritische Grenzen, die einer Zwischenauswertung Rechnung tragen, werden mittels fundamentaler Regeln der Wahrscheinlichkeitsrechnung abgeleitet.Schlussfolgerung: Mithilfe eines Wahrscheinlichkeitsbaums können die Prinzipien flexibler Studiendesigns einfach kommuniziert werden und somit hoffentlich zu guter Planung und Durchführung von klinischen Studien beitragen. Background: Flexible design of clinical trials may allow tailoring ongoing investigations to better suit actual circumstances such as an unexpectedly low effect size without compromising error control properties. Available statistical methods are, however, complex.Method: A two-stage flexible design is represented in form of a simple chance tree. Critical limits accounting for one interim analysis are derived using only basic rules of probability calculus.Conclusion: By means of chance trees the principles of flexible trial design may be communicated more easily, thus hopefully adding to good design and conduct of clinical trials. IntroductionPlanned interim analyses of clinical trials may help to assess whether ongoing trials can realistically be expected to answer its primary question(s). Flexible or adaptive designs have been advocated as a means to straighten out specific design errors such as an unexpectedly high variability and low effect size, thus possibly rescuing hampered experiments and accelerating the development of effective interventions . Regulatory guidance on the proper/improper use of flexible designs in drug development, e.g. with seamless phases II and III, is currently being compiled . By means of a simple yet novel chance tree I aim to make the fundamentals of flexible trial design, particularly (multiplicity) adjustment of critical limits for interim analysis, accessible to a broad audience. Meta-analysis within the same trialThe basic idea of adaptive design is to decompose a clinical trial into separate stages which results are then recombined in a within-trial meta-analysis . Various statistical methods are available for decision making, effect estimation and calculation of confidence intervals including, for example, combination of the p-values from the separate trial stages by Fisher’s combination test or the inverse normal method . An alternative, very general formulation is based on the conditional error function α(p1) which specifies the amount of conditional type I error to be spent for the second trial stage given the p-value p1 from the first stage of the trial , . Unmasked results from precedent stages may serve to optimize the design/power of subsequent stages without compromising the type I error control. Popular group-sequential plans, e.g. employing O’Brien-Fleming or Pocock stopping boundaries, are special instances of flexible design albeit without any major design modification such as adaptation of the maximum sample size, selection of a subpopulation, dropping treatment arms or changing endpoints/ hypotheses. Two-stage designs based on Fisher’s combination testAssume a single one-sided null hypothesis H0, e.g. treatment A is not superior to treatment B, is to be tested at level</PlainText></TextGroup> α in a two-stage design. Let p<Subscript>1</Subscript> and p<Subscript>2</Subscript> be the corresponding one-sided p-values from the separate, independent samples of the first and second stage, respectively. Boundaries α<Subscript>1</Subscript> and α<Subscript>0</Subscript> with α<Subscript>1</Subscript><α<α<Subscript>0</Subscript> for early stopping with rejection or acceptance of H<Subscript>0</Subscript>, respectively, need to be fixed in the trial protocol. The choice of α<Subscript>2</Subscript>≤α determines the α-spending between the interim analysis (local level α<Subscript>1</Subscript>) and the final combination test (local level α<Subscript>2</Subscript>) <TextLink reference="7"></TextLink>, <TextLink reference="8"></TextLink>. To obtain a level-α test procedure the quantities α, α<Subscript>0</Subscript>, α<Subscript>1</Subscript>, α<Subscript>2</Subscript> and the critical limit <ImgLink imgNo="1" imgType="inlineFigure"/>, where <ImgLink imgNo="2" imgType="inlineFigure"/> denotes the (1–α) quantile of the central χ<Superscript>2</Superscript>-distribution with 4 degrees of freedom for the final combination test, must fulfil the constraint <ImgLink imgNo="3" imgType="inlineFigure"/>, i.e. α<Subscript>1</Subscript> is iteratively determined given α and α<Subscript>2</Subscript>. For example, for α<Subscript>0</Subscript>=0.5 and α<Subscript>2</Subscript>=α=0.025, c<Subscript>0.025</Subscript>=0.0038 and α<Subscript>1</Subscript>=0.0102. This procedure can also be defined in terms of the conditional error function</Pgraph><Pgraph><Indentation><Indentation><ImgLink imgNo="4" imgType="inlineFigure"/></Indentation></Indentation></Pgraph><Pgraph>where <ImgLink imgNo="5" imgType="inlineFigure"/> (by definition) and H<Subscript>0</Subscript> can be rejected after the second stage if p<Subscript>2</Subscript>≤α(p<Subscript>1</Subscript>) <TextLink reference="9"></TextLink>.</Pgraph></TextBlock> <TextBlock linked="yes" name="Chance tree of a two-stage adaptive design"> <MainHeadline>Chance tree of a two-stage adaptive design</MainHeadline><Pgraph>A chance tree is a tree-like graph of chance event outcomes (so-called “nodes”). The probabilities of all directly successive nodes to a specifc node sum to one. The probability of a specific branch of the tree equals the product of all probabilities along this branch, and the probability of a bundle of branches equals the sum of all probabilities of branches contained.</Pgraph><Pgraph>A chance tree which under the null hypothesis approximates a simple two-stage adaptive design based on Fisher’s combination test is shown in Figure 1 <ImgLink imgNo="1" imgType="figure"/> <TextLink reference="3"></TextLink>. The number and position of intervals for p<Subscript>1</Subscript> (i.e. the nodes) were chosen to approximate the continuous decision boundary in a simple illustrative way. The boundaries of both adaptive tests are depicted in Figure 2 <ImgLink imgNo="2" imgType="figure"/>. The one-sided type I error α to be spread over all branches is 0.025. To approximate the smooth curve <ImgLink imgNo="6" imgType="inlineFigure"/> over the interval ]x<Subscript>1</Subscript>,x<Subscript>2</Subscript>] by a step function, the partial area under the curve was calculated by <ImgLink imgNo="7" imgType="inlineFigure"/> and then divided by x<Subscript>2</Subscript>–x<Subscript>1</Subscript>. If the one-sided p-value p<Subscript>1</Subscript> of the first stage is greater than α<Subscript>0</Subscript>=0.5 the trial is stopped. If p<Subscript>1</Subscript> is less or equal α<Subscript>1</Subscript>=0.0102 the null hypothesis is rejected and the trial is stopped. If p<Subscript>1</Subscript> is intermediate, i.e. less or equal α<Subscript>0</Subscript>=0.5 and greater than α<Subscript>1</Subscript>=0.0102, a second stage may be designed and conducted yielding the one-sided p-value p<Subscript>2</Subscript>. If p<Subscript>2</Subscript> is less than or equal the type I error “conditional on p<Subscript>1</Subscript>” the null hypothesis is rejected, and accepted otherwise. The conditional error is the (constant) density of α in the interval containing p<Subscript>1</Subscript>, e.g. 0.1055≈0.0026/(0.05–0.025) (error due to rounding) where 0.0026 is the area under the conditional error function over the interval ]0.025, 0.05]. Thus if a p-value p<Subscript>1</Subscript> between 0.025 and 0.05 is obtained at the first stage, the null hypothesis can be rejected at the second stage if p<Subscript>2</Subscript> is less or equal 0.1055 else it is retained. The other first stage nodes, i.e. the intervals ]0.35, 0.5], ]0.2, 0.35], ]0.1, 0.2] and ]0.0102, 0.025], are dealt with analogously. Note, each combination of a p<Subscript>1</Subscript>-branching (i.e. a partition of the interval [0, 1]) with a distribution of α over branches defines a different adaptive test. A similar chance tree can be grown to illustrate the (conditional) power of the adaptive procedure in any specific setting.</Pgraph><Pgraph>In a more precise approach, nodes (i.e. their number and position) may be chosen to keep the approximation error below any absolute (or relative) bound. Conditional on a fixed number of nodes this error may also be minimized over position (i.e. location and width of intervals). In the same way other p-value combination functions ρ can be approximated such as the inverse normal method defined by </Pgraph><Pgraph><Indentation><Indentation> <ImgLink imgNo="8" imgType="inlineFigure"/></Indentation></Indentation></Pgraph><Pgraph>(Φ<Superscript>–1</Superscript> denotes the inverse of the standard normal distribution function) with corresponding conditional error function </Pgraph><Pgraph><Indentation><Indentation> <ImgLink imgNo="9" imgType="inlineFigure"/></Indentation></Indentation></Pgraph><Pgraph>where u is the one-sided critical value of the two-stage group sequential test with provision for early stopping in favour of H<Subscript>0</Subscript> <TextLink reference="3"></TextLink>, <TextLink reference="4"></TextLink>, <TextLink reference="5"></TextLink>.</Pgraph><Pgraph></Pgraph></TextBlock> <TextBlock linked="yes" name="Conclusions"> <MainHeadline>Conclusions</MainHeadline><Pgraph>Flexible designs can be represented in form of a chance tree which is an intuitive mathematical tool most clinicians are familiar with from decision analysis. Thus, the principles of adaptive testing may be explained and communicated more easily, hopefully adding to good design and conduct of biomedical investigations.</Pgraph></TextBlock> <TextBlock linked="yes" name="Notes"> <MainHeadline>Notes</MainHeadline><SubHeadline>Conflicts of interest</SubHeadline><Pgraph>None declared.</Pgraph><SubHeadline>Acknowledgment</SubHeadline><Pgraph>The author thanks Walter Lehmacher, Gernot Wassmer and two anonymous reviewers for comments which helped to improve the presentation of the paper.</Pgraph></TextBlock> <References linked="yes"> <Reference refNo="1"> <RefAuthor>O'Neill RT</RefAuthor> <RefTitle>FDA's critical path initiative: a perspective on contributions of biostatistics</RefTitle> <RefYear>2006</RefYear> <RefJournal>Biom J</RefJournal> <RefPage>559-64</RefPage> <RefTotal>O'Neill RT. FDA's critical path initiative: a perspective on contributions of biostatistics. Biom J. 2006;48(4):559-64. DOI: 10.1002/bimj.200510237</RefTotal> <RefLink>http://dx.doi.org/10.1002/bimj.200510237</RefLink> </Reference> <Reference refNo="2"> <RefAuthor>Committee for Medicinal Products for Human Use</RefAuthor> <RefTitle></RefTitle> <RefYear>2006</RefYear> <RefBookTitle>Reflection paper on methodological issues in confirmatory clinical trials with flexible design and analysis plan</RefBookTitle> <RefPage></RefPage> <RefTotal>Committee for Medicinal Products for Human Use. Reflection paper on methodological issues in confirmatory clinical trials with flexible design and analysis plan. London: European Medicines Agency; 2006. Doc. Ref. CHMP/EWP/2459/02. Available from: http://www.emea.europa.eu/pdfs/human/ewp/245902en.pdf</RefTotal> <RefLink>http://www.emea.europa.eu/pdfs/human/ewp/245902en.pdf</RefLink> </Reference> <Reference refNo="3"> <RefAuthor>Bauer P</RefAuthor> <RefAuthor>Köhne K</RefAuthor> <RefTitle>Evaluation of experiments with adaptive interim analyses</RefTitle> <RefYear>1994</RefYear> <RefJournal>Biometrics</RefJournal> <RefPage>1029-41</RefPage> <RefTotal>Bauer P, Köhne K. Evaluation of experiments with adaptive interim analyses. Biometrics. 1994;50(4):1029-41. Correction: Biometrics. 1996;52:380. DOI: 10.2307/2533441</RefTotal> <RefLink>http://dx.doi.org/10.2307/2533441</RefLink> </Reference> <Reference refNo="4"> <RefAuthor>Lehmacher W</RefAuthor> <RefAuthor>Wassmer G</RefAuthor> <RefTitle>Adaptive sample size calculations in group sequential trials</RefTitle> <RefYear>1999</RefYear> <RefJournal>Biometrics</RefJournal> <RefPage>1286-90</RefPage> <RefTotal>Lehmacher W, Wassmer G. Adaptive sample size calculations in group sequential trials. Biometrics. 1999;55(4):1286-90. DOI: 10.1111/j.0006-341X.1999.01286.x</RefTotal> <RefLink>http://dx.doi.org/10.1111/j.0006-341X.1999.01286.x</RefLink> </Reference> <Reference refNo="5"> <RefAuthor>Proschan MA</RefAuthor> <RefAuthor>Hunsberger SA</RefAuthor> <RefTitle>Designed extension of studies based on conditional power</RefTitle> <RefYear>1995</RefYear> <RefJournal>Biometrics</RefJournal> <RefPage>1315-24</RefPage> <RefTotal>Proschan MA, Hunsberger SA. Designed extension of studies based on conditional power. Biometrics. 1995;51(4):1315-24. DOI: 10.2307/2533262</RefTotal> <RefLink>http://dx.doi.org/10.2307/2533262</RefLink> </Reference> <Reference refNo="6"> <RefAuthor>Müller HH</RefAuthor> <RefAuthor>Schäfer H</RefAuthor> <RefTitle>Adaptive group sequential designs for clinical trials: combining the advantages of adaptive and of classical group sequential approaches</RefTitle> <RefYear>2001</RefYear> <RefJournal>Biometrics</RefJournal> <RefPage>886-91</RefPage> <RefTotal>Müller HH, Schäfer H. Adaptive group sequential designs for clinical trials: combining the advantages of adaptive and of classical group sequential approaches. Biometrics. 2001;57(3):886-91. DOI: 10.1111/j.0006-341X.2001.00886.x</RefTotal> <RefLink>http://dx.doi.org/10.1111/j.0006-341X.2001.00886.x</RefLink> </Reference> <Reference refNo="9"> <RefAuthor>Posch M</RefAuthor> <RefAuthor>Bauer P</RefAuthor> <RefTitle>Adaptive two stage designs and the conditional error function</RefTitle> <RefYear>1999</RefYear> <RefJournal>Biom J</RefJournal> <RefPage>689-96</RefPage> <RefTotal>Posch M, Bauer P. Adaptive two stage designs and the conditional error function. Biom J. 1999;41(6):689-96. DOI: 10.1002/(SICI)1521-4036(199910)41:6<689::AID-BIMJ689>3.0.CO;2-P</RefTotal> <RefLink>http://dx.doi.org/10.1002/(SICI)1521-4036(199910)41:6<689::AID-BIMJ689>3.0.CO;2-P</RefLink> </Reference> <Reference refNo="7"> <RefAuthor>Bauer P</RefAuthor> <RefAuthor>Röhmel J</RefAuthor> <RefTitle>An adaptive method for establishing a dose-response relationship</RefTitle> <RefYear>1995</RefYear> <RefJournal>Stat Med</RefJournal> <RefPage>1595-607</RefPage> <RefTotal>Bauer P, Röhmel J. An adaptive method for establishing a dose-response relationship. Stat Med. 1995;14(14):1595-607. DOI: 10.1002/sim.4780141410</RefTotal> <RefLink>http://dx.doi.org/10.1002/sim.4780141410</RefLink> </Reference> <Reference refNo="8"> <RefAuthor>Bauer P</RefAuthor> <RefAuthor>Kieser M</RefAuthor> <RefTitle>Combining different phases in the development of medical treatments within in a single trial</RefTitle> <RefYear>1999</RefYear> <RefJournal>Stat Med</RefJournal> <RefPage>1833-48</RefPage> <RefTotal>Bauer P, Kieser M. Combining different phases in the development of medical treatments within in a single trial. Stat Med. 1999;18(14):1833-48. DOI: 10.1002/(SICI)1097-0258(19990730)18:14<1833::AID-SIM221>3.0.CO;2-3</RefTotal> <RefLink>http://dx.doi.org/10.1002/(SICI)1097-0258(19990730)18:14<1833::AID-SIM221>3.0.CO;2-3</RefLink> </Reference> </References> <Media> <Tables> <NoOfTables>0</NoOfTables> </Tables> <Figures> <Figure format="png" height="627" width="635"> <MediaNo>1</MediaNo> <MediaID>1</MediaID> <Caption><Pgraph><Mark1>Figure 1: A chance tree of a two-stage adaptive test</Mark1></Pgraph></Caption> </Figure> <Figure format="png" height="534" width="528"> <MediaNo>2</MediaNo> <MediaID>2</MediaID> <Caption><Pgraph><Mark1>Figure 2: Decision boundaries of a two-stage adaptive Bauer-Köhne test (green solid line) and its chance tree approximation (blue broken line). Any pair of p-values (p</Mark1><Mark1><Subscript>1</Subscript></Mark1><Mark1>, p</Mark1><Mark1><Subscript>2</Subscript></Mark1><Mark1>) above [on or under] the respective boundary leads to acceptance [rejectance] of the null hypothesis. </Mark1></Pgraph></Caption> </Figure> <NoOfPictures>2</NoOfPictures> </Figures> <InlineFigures> <Figure format="png" height="18" width="164"> <MediaNo>1</MediaNo> <MediaID>1</MediaID> <AltText>Equation 1</AltText> </Figure> <Figure format="png" height="18" width="62"> <MediaNo>2</MediaNo> <MediaID>2</MediaID> <AltText>Equation 2</AltText> </Figure> <Figure format="png" height="18" width="145"> <MediaNo>3</MediaNo> <MediaID>3</MediaID> <AltText>Equation 3</AltText> </Figure> <Figure format="png" height="81" width="229"> <MediaNo>4</MediaNo> <MediaID>4</MediaID> <AltText>Equation 4</AltText> </Figure> <Figure format="png" height="29" width="99"> <MediaNo>5</MediaNo> <MediaID>5</MediaID> <AltText>Equation 5</AltText> </Figure> <Figure format="png" height="16" width="67"> <MediaNo>6</MediaNo> <MediaID>6</MediaID> <AltText>Equation 6</AltText> </Figure> <Figure format="png" height="19" width="117"> <MediaNo>7</MediaNo> <MediaID>7</MediaID> <AltText>Equation 7</AltText> </Figure> <Figure format="png" height="20" width="211"> <MediaNo>8</MediaNo> <MediaID>8</MediaID> <AltText>Equation 8</AltText> </Figure> <Figure format="png" height="78" width="374"> <MediaNo>9</MediaNo> <MediaID>9</MediaID> <AltText>Equation 9</AltText> </Figure> <NoOfPictures>9</NoOfPictures> </InlineFigures> <Attachments> <NoOfAttachments>0</NoOfAttachments> </Attachments> </Media> </OrigData> </GmsArticle>