Info

Negative

Positive

Off study

Randomize

Randomize

Experimental Control

Experimental Control

Fig. 1. Targeted clinical trial design for evaluating a new experimental therapy. A biomarker classifier is developed for identifying those patients most likely to respond to the new treatment (E). Only those patients are randomized to E versus the control treatment. The patients predicted less likely to respond (marker negative) are off-study. The targeted design is most useful in cases where the biomarker classifier has a strong biological rationale for identifying responsive patients and where it may not be ethically advisable to expose marker negative patients to the new treatment.

For many molecularly targeted drugs, however, the appropriate assay for selecting patients is not known and development of a classifier based on comparing expression profiles for phase II responders versus phase II non-responders may be the best approach. In such instances, one may not have sufficient confidence in the genomic classifier developed in phase II to use it for excluding patients in phase III trials as in Fig. 1. It may be better in this case to accept all conventionally eligible patients, and use the classifier in the pre-defined analysis plan.

Figure 2 shows the marker by treatment interaction design discussed by Sargent et al. (18) and by Pusztai and Hess (19). Both marker positive and marker negative patients are randomized to the experimental treatment or control. The analysis plan either calls for separate evaluation of the treatment difference in the two-marker strata or for testing the hypothesis that the treatment effect is the same in both marker strata.

When this design is used for development of an experimental drug, an appropriate analysis plan might be to utilize a preliminary test of interaction; if the interaction is

Approximate Number of Events Required for 80% Power with 5% Two-Sided Log-Rank Test for Comparing Randomized Arms of Design Shown in Fig. 1. Only Marker + Patients Are Randomized. Treatment Hazard Ratio for Marker + Patients Is Shown in First Column. Time-To-Event

Distributions Are Exponential

Table 1

Hazard Ratio for Marker + Patients

Number of Events Required

66 196

Table 2

Approximate Number of Events Required for 80% Power with 5% Two-Sided Log-Rank Test for Comparing Treatment Versus Control arm of Design in which Marker is not Measured. Randomized Arms Are Mixtures of Marker - and Marker + Patients. Hazard Ratio For Marker -Patients Is 1 for the Two Treatment Groups and 0.67 For Marker + Patients.

Table 2

Approximate Number of Events Required for 80% Power with 5% Two-Sided Log-Rank Test for Comparing Treatment Versus Control arm of Design in which Marker is not Measured. Randomized Arms Are Mixtures of Marker - and Marker + Patients. Hazard Ratio For Marker -Patients Is 1 for the Two Treatment Groups and 0.67 For Marker + Patients.

% of Patients Marker

Approximate Number of Events Required

20

5200

33

1878

50

820

not significant at a pre-specified level, then the experimental treatment is compared to the control overall. If the interaction is significant, then the treatment is compared to the control within the two strata determined by the marker. The sample size planning for such a trial and determination of the appropriate significance level for the preliminary interaction test are discussed by Simon (20).

Simon and Wang (21) proposed an alternative analysis plan for the design of Fig. 2. They suggested that the overall null hypothesis for all randomized patients is tested at the 0.04 significance level. A portion, e.g., 0.02, of the usual 5 percent false positive rate is reserved for testing the new treatment in the subset predicted by the classifier to be responsive. The analysis starts with a test of the overall null hypothesis, without a preliminary test of interaction. If the overall null hypothesis is rejected, then one concludes that the treatment is effective for the randomized population as a whole and that the classifier is not needed. If the overall null hypothesis is not rejected at the 0.03 level, then a single subset analysis is conducted; comparing the experimental treatment to the control in the subset of patients predicted by the classifier as being most likely to be responsive to

Measure marker

Measure marker

Negative Positive

Negative Positive

Randomize

Randomize

Randomize

Randomize

Fig. 2. Stratified analysis design for evaluating a new experimental treatment (E) relative to a control (C). The status of a biomarker based classifier of the likelihood of responding to E is utilized in a prospectively specified analysis plan. The biomarker classifier is not just used for stratifying the randomization. Alternative analysis plans are described in the text.

the new treatment. If the null hypothesis is rejected, then the treatment is considered effective for the classifier determined subset. This analysis strategy provides sponsors an incentive for developing genomic classifiers for targeting therapy in a manner that does not unduly deprive them of the possibility of broad labeling indications when justified by the data.

10 Ways To Fight Off Cancer

10 Ways To Fight Off Cancer

Learning About 10 Ways Fight Off Cancer Can Have Amazing Benefits For Your Life The Best Tips On How To Keep This Killer At Bay Discovering that you or a loved one has cancer can be utterly terrifying. All the same, once you comprehend the causes of cancer and learn how to reverse those causes, you or your loved one may have more than a fighting chance of beating out cancer.

Get My Free Ebook


Post a comment