Caren van Roekel

276 Chapter 10 How should response to treatment be evaluated? First of all, there is a difference in response evaluation between medical products (i.e. pharmaceuticals) and medical devices. In patients with liver- dominant disease with a few enlarged lymph nodes, chemotherapy will have an effect on both the liver and on the lymph nodes. On the contrary, a medical device such as radioactive microspheres, is implanted in an organ and only has a local effect. In the case of radioembolization, this effect is on the liver lesions only. In the case of response evaluation on a whole-body level, patients who have complete response in the liver but growth of extrahepatic metastases may still be categorized as having progressive disease (98). In such cases, it may seem that treatment with radioembolization is ineffective, while it actually worked very well. Secondly, the efficacy of treatment can be measured in many ways: not only as local response, but also as progression-free survival (PFS) or overall survival (OS). In the case of patients with colorectal cancer metastases who are treated in the first line, PFS is a better endpoint than OS since patients often receive multiple subsequent lines of treatment. OS can still be used, but in that case, the efficacy of the combination of treatments that patients receive should be evaluated. There are several measures for response evaluation as a surrogate endpoint. The two most commonly used response metrics in the field of radioembolization are the Response Evaluation Criteria In Solid Tumors (RECIST) version 1.1 (98), and the Positron Emission Tomography Response Criteria In Solid Tumors (PERCIST) (99). RECIST is based on anatomical response, whereas PERCIST is based on metabolic response. In both metrics, response is categorized in complete response (CR), partial response (PR), stable disease (SD) or progressive disease (PD). A drawback of both methods, however, is that the definition of stable disease allows for a minor shrinkage or a minor growth of tumors, which may oversimplify the process (100). Another drawback of these methods is that target lesions (one per organ for PERCIST, a maximum of two per organ for RECIST 1.1) need to be defined. This is not always easy, because patients can have disseminated disease with confluent tumors. In the case of PET, if there is only one connecting voxel between two lesions, they may appear as one confluent lesion, rendering separate evaluation difficult. Furthermore, target