Patients' baseline characteristics with respect to age, sex, and tumor histology including differentiation and Child-Pugh score are shown in Table . Information to sorafenib doses at follow-up is included in Table .
Response Assessment by Measurement of Size and Volume
In 38% of our patients, tumor lesions experiencing morphological signs compatible with tumor necrosis under sorafenib showed at least a temporary increase in tumor size, up to 58% of the initial size of the lesion, suggesting that this might reflect an increased volume of liquid tumor parts. In 3 patients volume increase of HCC was found to be over the 20% threshold stipulated by RECIST, in 3 other patients volume increase was >10% of the initial lesion's volume (Table ). In 3 patients (#1, #13, and #15) a substantial reduction of the whole tumor volume could be achieved under sorafenib (reduction of 36%, 45%, and 30%, respectively). Of note, two patients (#7, #21) undergoing long- and short-term sorafenib dose interruptions experienced repeated responses with temporal variations in tumor volume.
Response Assessment by Measurement of Tumor Necrosis
On post-gadolinium fat-saturated T1WI, areas of tumor necrosis appeared de novo in 7/21 patients (Table ) and 13/39 lesions (data not shown) or increased in 9/21 patients (Table ) and 9/39 lesions (data not shown). In patients with progressive necrosis 2 lesions developed <25% necrosis, 4 lesions achieved <50% necrosis, and 7 lesions developed >75% tumor necrosis at follow-up. Altogether, progression of necrosis (↑) was diagnosed in 40/117 followed-up lesions (data not shown). Notably, reduction (↓) of the volume of tumor necrosis was registered only in 3 patients (#7, #18, and #21). The former two had progressive disease at follow-up with revitalization of tumor. The latter presented an undulant course of the disease due to intermittent dose discontinuations.
Assessment of T1WI and T2WI signal changes as a surrogate marker for intratumoral hemorrhage
A focal or diffuse increase in tumor signal to baseline, confirmed independently by both readers, was detected on nonenhanced T1WI in 15/21 patients (Table ) and 28/39 target lesions analysed (data not shown) at an early time point after a median of 5 weeks (range, 2–9 weeks) following onset of sorafenib therapy, and considered suspicious of hemorrhage and/or protein-rich necrosis. Interestingly, one patient with a complete response (#10) displayed early signal changes in T1WI, T2WI and an increase of necrotic areas within the tumor already 2 weeks after initiation of sorafenib therapy 400 mg bid. Increased signal on T1WI returned to normal with time, but the dynamics of this phenomenon could not be entirely analysed due to different time spans of the patients' individual MR-surveys. In five patients (#5, #11, #12, #14, #18), the T1WI tumor signal remained unchanged during follow-up. However, two of them developed an increase of the T2-signal (#5, #12) while in two patients a substantial progression of the necrotic tumor area could be detected (#12, #14) at time.
A synchronous signal increase on T2WI SE-sequences or rarely a temporary signal decrease was considered as further supporting the hypothesis of tumor hemorrhage. Almost synchronously with early signal intensity changes on T1WI, signal abnormalities on T2WI were found (median, 5.3 weeks; range, 2–12 weeks), consisting in most of the cases (15/21 patients) of an increase in signal intensity (Table ) and in 26/39 target lesions analysed (data not shown), whereas a decrease in T2-signal lesions was found in only 2/21 patients. In two patients (#5, #12) with stable signal patterns on T1WI, signal changes could be assessed only on T2WI. Thus, early signal abnormalities suggestive of tumor hemorrhage occurring after onset of sorafenib therapy were encountered in 15/21 patients.
Typical examples of signal intensity changes of MRI baseline imaging and follow-up investigations are presented in Figure , and for three representative patients (patients #7, #8, and #17). In Figure , typical changes from a patient presenting with a T1WI hypointense tumor signal at baseline and consecutive changes under therapy are shown. Figure demonstrates synchronous hemorrhagic necrosis in all hepatic lesions of a patient with multicentric HCC 3 weeks after onset of sorafenib therapy. Figure demonstrates increased signals on T1WI and T2WI occurring synchronously in all HCC manifestations after onset of sorafenib therapy. Of note, one patient (#21) presented a slow but continuous growth of the hepatic lesions despite typical signal changes on T1WI and T2WI. This patient had repeated therapy interruptions and a dose reduction due to a hand-foot syndrome, but a repeated increase in T1WI was observed following the re-onset of sorafenib. Two patients in this series, who showed a continuous tumor growth under treatment (#11, #18), showed no sorafenib related signal abnormalities on MRI at all and no de novo occurring or progressive necrosis.
Figure 1 A-F. 63-year-old female patient with multicentric HCC (patient #8). Axial nonenhanced T1WI performed at baseline showed a 10 cm large tumor (arrow) in the left liver lobe (Figure 1A). Note a tumor signal (arrow) slightly hypointense to normal liver parenchyma (more ...)
Figure 2 A-F. 71-year-old male patient with multifocal HCC (patient #9). Axial nonenhanced fat-suppressed T1WI imaging of the liver performed at baseline showed multiple mild hypointense HCC lesions (arrow) (Figure 2A). On corresponding coronal HASTE (T2-weighted) (more ...)
Figure 3 A-F. 61-year-old male patient with metastatic HCC (patient #17). Axial nonenhanced T1WI imaging of the liver showed hypointense signal of all HCC lesions (arrows) (Figure 3A). On corresponding T2WI images, HCC lesions demonstrated all a moderate hyperintense (more ...)
Potential Influence of Sorafenib Dose on Tumor Signal
Thus, early changes occurred in the 8 patients that were able to take sorafenib 800 mg daily within a mean auf 4.1 weeks (range, 3–5). Taking all 18 patients together who received either sorafenib 800 mg or 400 mg daily, the latter being the first dose reduction level, early changes could be detected within a mean of 4.2 weeks (range, 2–9), demonstrating that both dose levels had a similar reaction pattern. However, in a further patient who had to reduce sorafenib to 200 mg due to adverse reactions early in the therapy course, MRI signal alterations and new tumor necrosis could be detected at week 12 under therapy. Of note, the 2 patients with a longer duration until a signal response could be detected (#3, #16) had to temporarily discontinue therapy within the first two weeks due to a grade 3 skin reaction or a transient increase in bilirubin levels, respectively.
In conclusion, we show for the first time that rapid changes in MR signal intensities occur soon after onset of sorafenib therapy. A statistical analysis addressing the presence or absence of an increased T1WI signal, tumor necrosis, and increase in the necrotic volume in relation to the entire tumor volume [NV/TV] showed a significant relationship between T1WI and tumor necrosis (p = 0.017) and between tumor necrosis and NV/TV (p = 0.002). The direct association between T1WI and NV/TV did not reach statistical significance (p = 0.129) in this small sample size