|
Abstract:
|
Since the Clinton-Kessler initiative of 1994, the use of imaging in clinical trials has grown steadily along with the proliferation of image evaluation criteria. As images became better, so did the amount of information available to the image reviewers; yet with better images came a deluge of information that consequently resulted in more and more complex evaluation criteria that often include multiple imaging modalities in a composite rule set that has little consensus but little to no validation. Added to this are the novel therapies that defy analysis by the standard set of criteria and need to rely on more novel methods that will also need to be at least partially validated to submit to the regulators. Statistical dilemmas abound and adaptive biomarker studies attempt to address study-specific biomarkers but the complex flood of image information presents a challenge to the statistician to define what it is that is exactly being measured. This paper looks at the many different evaluation criteria involved in oncology imaging, presents statistical dilemmas faced by study teams in implementing these criteria, presents 2 case studies in the use of complex criteria and provides
|
