Instructive use-case problems for clinical trial imagers

Return to Main Page :::curator: carljaffe@alum.mit.edu

---

You are approached by a company that is developing a new molecular drug targeting tumor vascular growth for stage II-III lung cancer. They plan a large phase III trial for FDA registration and want you to advise them on developing a protocol for an imaging endpoint. Note that unlike in strictly scientific research, imaging serves a different purpose in FDA registration trials since here imaging, even though an end-point must be subordinate to the therapy test design in order to achieve rapid accrual and must also acknowledge the diverse capabilities of the multiple institutions involved.

• Would you recommend they use PFS or ORR http://ctwiki.rsna.org/index.php?title=Main_Page#Abbreviations_and_Definitions?

• Are there obvious precedents by FDA regarding trials that address specific diseases, or whether the therapeutic indication is intended for first line vs. second or third line use?

• Would you use RECIST or WHO measurement criteria?

• What should be the imaging intervals for most efficiency and least cost and what decision point should be used to define discontinuance of imaging?

• What primary imaging method would you recommend – CT, MR, PET?

• Would you make use of, or allow, other non-primary endpoint imaging to be included and if so what would you do with the data (e.g. a bone scan etc)?

• If you choose MR or CT would you require one or the other to be used exclusively or would you allow site convenience/preference to occur?

• If the latter, what body parts would you scan – thorax only, thorax to liver, thorax to pelvis?

• Would you scan the brain and/or bone at baseline, and if you did would you require it each time?

• How would you specify slice thickness and slice adjacency? Would it be a range/upper bound or would it be a specific thickness for all sites?

• Would CT kVp/mAs or MR pulse sequence have to be tightly specified or could it be defined liberally given that this is a large multi-site trial?

• Would you require contrast; and if so would you insist on acquisition of both non-contrast and contrast at each study time?

• How would you change these parameters if the cancer were pancreas or colon rather than lung?

• Would it be necessary to particularly specify consistency of scan direction such as caudal - to-cephalad or vice-versa (to minimize contrast dynamic variability)?

• If contrast were to be specified, would it be sufficient to define contrast amount and scan delay start time or should machine detected parameters such as contrast arrival time be specified? Or is that important only in the context of protocols that expect to use a contrast dynamic study as endpoint?

• Would you require central reads, and if so why? Should the readers be blinded as to which target lesions the other reader has chosen for measurement?

• Would it make any difference if the treatment arm couldn’t be blinded?

• Would it make any difference if the drug was a cytotoxic as opposed to a targeted drug?

• How do the FDA presentations on ‘image charters’ affect your image protocol planning

• Given that imaging is a primary endpoint does it matter what the statistical powering target is (e.g. an 80% or 90% difference to avoid a type I error)

• What role could/should FDA or NCI play to help you develop such a protocol?