Targeted therapy refers to a new generation of anticancer drugs, designed to interfere with distinct molecular entities, involved in tumor growth and/or progression. The multikinase inhibitor sorafenib demonstrated profound anti-tumor activity in different clinical settings, including HCC [5, 6, 17], renal cell carcinoma (RCC) [9, 18], as well as other advanced, so far therapy-refractory tumor entities [19, 20]. In the special situation of HCC, imaging studies have to take into account that tumor areas develop within a heavily diseased liver, in which several clonal tumor areas with different patterns of differentiation might exist simultaneously [2, 3].
In our patient population a substantial impact of sorafenib on the MRI pattern of HCC tissues was observed, which might be regarded at least as some kind of a therapy response; however, the majority of these changes did not lead to a classification of a partial response according to classical RECIST criteria. Nevertheless, 16 out of 21 patients presented with either new or progressive necrosis in repeated MRI scans, which suggests that quantifying necrotic areas in relation to the entire tumor volume might be useful as a reliable predictive marker of therapy response. Similar results have been described for 11 HCC patients under sorafenib treatment using a semiautomated computerized technique to analyse tumor necrosis in contrast enhanced CT scans . However, the influence of tumor necrosis on the clinical outcome, including overall survival data, has not been prospectively investigated yet.
Reaction patterns to targeted agents like sorafenib seem to foremost include (i) disease stabilization (rather than a direct cytotoxic effect accompanied by tumor shrinkage) or (ii) an induction of intralesional necrosis that does not automatically lead to a marked decrease in tumor size. In this context, a surprising finding of our study was that more than one third of the tumors that displayed new necrotic areas under treatment showed a temporary expansion, which seemed to be primarily due to an increased volume induced by tumor necrosis and not by an accumulation of vital tumor cells. Interestingly, comparable data have been reported in patients with gastrointestinal stromal tumors (GIST), a different tissue context receiving therapy with the tyrosine kinase inhibitor imatinib mesylate (STI571; Gleevec™) . This information must be taken into account when judging tumor responses under treatment with both, single kinase (e.g. imatinib mesylate) as well as multikinase inhibitors (e.g. sorafenib).
Especially in the context of HCC, MRI provides a highly sensitive method for detecting soft tissue signal changes, to assess the extent of therapy-related tumor necrosis, and to monitor both, distinct hepatoma nodules as well as the surrounding liver parenchyma in exquisite detail . It is obvious that signal abnormalities related to sorafenib that occur at a distinct time point are only temporary in nature because of a continuous alteration of therapy induced signal abnormalities, e.g. by physiologic processes, leading to hemoglobin degradation or change in the tissue protein content. Both protein-rich secretion due to rapid tumor necrosis or hemorrhage could explain signal abnormalities on MRI. However, taken all signal abnormalities occurring on T1WI and T2WI and their temporary evolution into account, hemorrhage seems to be much more plausible. Theoretically, looking at an early time point soon after the initiation of a targeted therapy, the appearance of intracellular deoxyhemoglobin is expected to induce low signal intensities on both T1WI and T2WI, similar to well-known signal abnormalities, occurring early in acute hemorrhage due to other causes (days 1 – 3 after occurrence). However, detection of such changes is expected to be less sensitive in tumors that present initially with either isointense signals or signal intensities that are only slightly different from that of adjacent liver parenchyma and would require a very early monitoring, for instance by MR-imaging, which was not the focus of our study. In the subsequent early subacute phase (> 3 days after onset of hemorrhage), the transformation of intracellular deoxyhemoglobin to methemoglobin is known to induce a change of the MR-signal to high intensities on T1WI and low intensities on T2WI images, whereas in the late subacute phase (> 7 days post hemorrhage), the occurrence of extracellular methemoglobin results in high signals in both T1WI and T2WI, which contrasts much better with the native tumor and liver parenchymal signal. This time span (first weeks after initiation of therapy) seems to be most adequate for accurate demonstration of hemorrhage in tumors as caused by sorafenib and corresponds to our imaging schedule. Other theoretical causes for an increase of signal intensity on nonenhanced T1WI include a protein-rich necrosis, similar to a coagulation necrosis that occurs after radiofrequency ablation. However, in that setting, the T2WI signal of those lesions is expected to be low and persists for a much longer time period than hemorrhagic alterations observed in our sorafenib treated patients. Nevertheless, the exact time required for the reconversion of MR signals to baseline values could not be reliably determined, as the influence of variable sorafenib dosages remains unknown.
Alternative explanations for signal abnormalities occurring during sorafenib treatment in our cohort include tumor hemorrhage or necrosis induced by tumor progression. However, the latter could indirectly be excluded by further follow-up MR imaging studies, which demonstrated either tumor size stabilization or more often a continuously rising ratio of the volume of necrotic tumor areas to viable tumor tissue.
There are some limitations of our work, first of all caused by the small number of patients recruited and evaluated. Follow-up studies with larger numbers of patients are needed in order to validate our results and establish the accuracy of such a response assessment. Secondly, for scientific reasons it would be desirable to compare radiological signal changes under therapy with histological alterations. However, due to ethical considerations, such a parallel evaluation is not feasible in humans. Nevertheless, long-term monitoring as in our study is regarded as an acceptable alternative.