Get documents about "Increasing efficiency in HIV drug development methods for stopping
Document Sample
5th IAS, 19-22 July, Cape Town, South Africa
Increasing efficiency in HIV drug development: Abstract: CDB070
methods for stopping futile studies early
Sara Hughes1, Robert L Cuffe1, Alfons Lieftucht2, and W Garrett Nichols3
1Infectious Diseases Statistics, 2Decision Sciences, 3Infectious Diseases Medicines Development Centre, GSK R&D
Corresponding author: sara.h.hughes@gsk.com
Tools Developed (as applied to efficacy endpoint
Introduction for case study)
Tool 4. Illustration of decision-analytic tool in use:
Comparing predicted efficacy outcomes for 80% vs 70% POS thresholds
As the safety and efficacy of available ARTs improve, requirements enabling Tool 1. Balancing risk of incorrectly stopping a succeeding study with
approval of new drugs become more rigorous. For noninferiority trials, the risk of allowing a failing study to continue 80% POS threshold 70% POS threshold
progressively raising the data bar may result in prohibitively large trials,
potentially stifling drug development [Hill, 2008]. Futility designs enable the early
termination of trials where initial data indicate the drug has little chance of Setting a stopping
demonstrating study objectives. However, limited guidance exists for selecting
chance of false stop (%)
study continuation criteria at a futility interim analysis that balance the risk of
threshold of 70% POS
stopping a succeeding study versus allowing a failing study to continue. will lead to a 27%
chance of incorrectly
Methods stopping at the interim
and a 10% chance of
We developed novel graphical and decision-analysis tools to enable optimal
selection of continuation criteria at futility interim analyses [Hughes, 2009]. incorrectly continuing
Graphs developed quantified, for each possible threshold: the study
• the risk of incorrectly stopping a succeeding study; Stop at interim Continue to good results
• the risk of allowing a failing study to continue; Continue to failed study Continue to excellent results
• how these risks vary depending on the timing of the interim analysis; Continue to mediocre results
• the reduction in the study’s power by including a futility assessment. Results
Decision analysis incorporated drug performance expectations and allowed chance of false go (%)
The stopping thresholds chosen for this case study (as detailed under Figure 1)
consideration of the incremental effect of relaxing thresholds upon the probability The tool above allows the user to compare and contrast the two key risks were calculated to have 80% chance of stopping a failing study while only
of continuation and success. associated with futility designs. The relative importance of these two risks is reducing the trial’s absolute probability of success by 9%. At the case-study’s
expected to vary according to the specifics of the trial in question. For example, futility assessment, criteria for study continuation were not met. Although the
These tools were then applied in a recent HIV case-study [Carosi, ICAAC 2008]. in late phase trials of alternative dosing regimens, higher stopping thresholds are experimental regimen performed well virologically, the stop-decision was
The group-sequential design formally assessed futility after recruiting ~25% of likely to be more appropriate. In contrast, for early phase trials with new drugs, it confirmed as appropriate since the hypothesized safety advantages were not
728 subjects; if continuation criteria were met, enrolment of the remaining is likely that lower stopping thresholds would be appropriate to terminate only a demonstrated. Thus the tools developed resulted in successful application,
subjects would resume. Details of the design are shown in Figure 1 below. completely ineffective drug or dose. redirecting 528 subjects and research funds to more productive trials.
Figure 1. Example study design Tool 2. Impact of when the futility
assessment occurs
Tool 3. Impact of futility assessment
on power for primary endpoint Conclusions
Note: In this example, continuation guidelines were not met at futility Despite their great potential, futility designs have been under-utilised to date. Of
assessment, and study did not proceed to Stage Two (shaded area) the trials reviewed by Hill, 28% (enrolling 2881 subjects) were unsuccessful in
demonstrating non-inferiority. Substantial subject time and resources could be
chance of false stop (%)
saved by terminating these studies early.
Stage One (N=200)
24 Week Futility Interim Analysis*
Power (%) A variety of statistical stopping methods can be used, but the choice of stopping
Stage One Subjects threshold is more important than choice of method. However, selection of the
continue in study for 48
FPV 1400mg QD
Weeks
optimal stopping threshold is challenging and there is a lack of published
RTV 100mg QD
ABC/3TC FDC QD practical guidance. Graphical and decision analytic tools as shown above enable
HIV+
ARV-naïve
the discussion and selection of optimal futility thresholds.
VL ≥1,000 c/mL
Drawbacks to using futility stopping rules include the increased risk of stopping a
Stage Two (N=528)
FPV 700mg BID successful drug or dose compared to having no futility assessment in a study.
RTV 100mg BID
ABC/3TC FDC QD FPV 1400mg QD However, this drawback should be considered in the context of minimising patient
RTV 100mg QD
subjects recruited POS threshold exposure to failing drugs/doses. Ultimately, futility designs are useful tools that
ABC/3TC FDC QD
Tool 2 illustrates how the risks associated with futility designs decrease, the later enable efficient testing of clinical and scientific hypotheses in drug development.
References
FPV 700mg BID the futility assessment is performed. This then enables the user to assess risk
RTV 100mg BID
ABC/3TC FDC QD versus likely benefit according to timing of analysis in the trial. Tool 3 allows the
user to investigate the resultant loss in trial power depending on the stopping Hill A and Sabin C. Designing and interpreting HIV noninferiority trials in naive and experienced patients.
AIDS 2008; 22: 913–921.
threshold selected. The graphs above provide clear illustration of the risks and Hughes S, Cuffe RL, Lieftucht A, Nichols WG. Informing the selection of futility stopping thresholds: case
*Futility assessment stopping thresholds based on key study objectives:
benefits of varying stopping thresholds. However, in order to quantitatively include study from a late-phase clinical trial. Pharmaceutical Statistics 2009; 8 (1): 25-37.
STOP if probability of success (POS) to demonstrate non-inferiority for proportion <400 c/mL (MD=F, Carosi G et al. Efficacy and Safety of Fosamprenavir + Ritonavir (FPV/RTV) 700mg/100mg Twice Daily
ITT-E) <70%. [Note: probability of success calculated using conditional power under the current trend.]
already available information on the new regimen’s expected performance (from
(BID) Versus FPV/RTV 1400mg/100mg Once Daily (QD) with ABC/3TC QD over 24 Weeks. ICAAC 2008,
PK and pilot studies), decision analysis can be used (Tool 4). This can also factor Washington DC, Abstract H-1244.
Or, STOP if POS to demonstrate 13 mg/dL difference in mean change from baseline in fasting non-HDL
cholesterol (ITT-E) <60% in expectations about the likely clinical acceptance of study results.
Created by
Related docs
