Targeted therapies in renal cell cancer: recent developments in imaging

Assessment of morphological changes, i.e., changes in tumor size burden, is an important evaluation method to determine the effect of anticancer drugs. Currently, RECIST is the most commonly used measurement system [23]. RECIST is a one-dimensional method and is based on the sum of the longest diameters of the appointed target lesions in the transversal plane. To determine tumor response accurately, target lesions (≥10 mm) are usually identified on CT or magnetic resonance imaging (MRI). According to RECIST, chest X-ray may also be considered instead of CT if lesions are clearly defined, have a diameter of at least 20 mm, and are surrounded by aerated lung. However, chest CT is preferred over chest X-ray, as CT is more sensitive than X-ray, especially in identifying new lesions.

RECIST provides four classes to determine the objective tumor response for target lesions (Table 1). Briefly, complete response (CR) is defined as disappearance of all target lesions, whereas a partial response (PR) requires a decrease in the sum of diameters of at least 30%, taking as reference the sum of the diameters at baseline. The appearance of new lesions or at least a 20% increase in the sum of the diameters is classified as PD, taking as reference the smallest sum during treatment. When the changes in the sum of the diameters are not sufficient to qualify for PR or PD, the objective response is defined as stable disease (SD).

Recently, RECIST has been updated and version 1.1 has been introduced in the clinic [23]. As compared to the old version 1.0 [25], notable changes include a reduction in the number of target lesions to be assessed per patient (from a maximum of 10 to a maximum of 5) and per organ (from 5 to 2 per organ), a change in assessment of pathological lymph nodes (measurement of the longest axis has been changed into the shortest; nodes should have a short axis of ≥15 mm to be considered as measurable) and the clarification of disease progression in small tumor burden (a 5 mm absolute increase is now required).

Many studies have applied RECIST to determine response to targeted therapy in mRCC [7, 8, 18]. Best response may take several months [8, 13], and is occasionally achieved up to 10 months after the start of treatment [22]. In a previous study in 55 mRCC patients treated with sunitinib, 13% of patients had PR, 69% of patients had SD, and 18% of patients had PD at a median of 1.9 months after the start, but 18% of patients with initial SD reached a PR at later time-points (median time to PR: 3.9 months; range: 2.4–9.7 months) [22]. Therefore, with the use of RECIST at first evaluation in patients with SD it cannot be discriminated which patients will have PD at the next time-point in order to timely change the type of treatment, since some patients benefit from continued SD and some patients may even reach a PR.

Since the expected effect of targeted agents in mRCC would be stabilization rather than substantial tumor regression, PR according to RECIST may not be accurate in the era of targeted therapy. Therefore, Thiam et al. [26] have attempted to determine a threshold for CT evaluation that best reflected treatment outcome. With PFS as the primary outcome, thresholds from −45% to +10% were tested in 334 mRCC patients treated with sunitinib. A decrease of at least 10% in the sum of the longest diameters was identified as the most accurate threshold to distinguish responders (n = 256) from non-responders (n = 78). Already during the first treatment cycle, 73% of patients reached a response according to the −10% threshold, whereas only 19% of patients reached a PR according to RECIST (≥30% decrease). After the second treatment cycle, the −10% and −30% thresholds were achieved in 93% en 64% of the patients. This study shows that responders can be earlier identified with the −10% threshold than with the conventional threshold of −30% according to RECIST.

Impressive changes in tumor attenuation have been observed during targeted therapy as visualized by contrast-enhanced CT. The contrast-enhancement on CT is directly related to tumor angiogenesis [27‐29]. In metastases from RCC, higher pretreatment values of contrast-enhancement are associated with a higher response rate to targeted therapy consisting of either sunitinib or sorafenib [30], allowing prediction of outcomes of individual metastatic lesions before the start of targeted therapy. In addition, treatment-induced changes in tumor density are associated with response to therapy and the development of necrotic areas [31]. In contrast, the development of a marked central fill-in and new enhancement in homogeneously, hypo-attenuating, non-enhancing RCC lesions are commonly seen at or just before the time of progression late in the course of treatment [32], suggesting that an increase in contrast-enhancement may be associated with disease progression. Currently, three different response evaluation criteria that incorporate changes in contrast-enhancement have been investigated in mRCC patients: the Choi criteria, the modified Choi criteria, and the size and attenuation CT (SACT) criteria (Table 1).

Using contrast-enhanced CT scans, Choi et al. [33, 34] have developed new evaluation criteria to determine the efficacy of imatinib in patients with gastrointestinal stromal cell tumors (GIST). In patients with GIST, imatinib is known to induce extensive tumor necrosis that may be accompanied by an increase in tumor size and may even simulate PD [34]. The Choi criteria include changes in tumor attenuation expressed as Hounsfield Units (HU) on contrast-enhanced CT. The threshold for changes in tumor attenuation on CT is based on a decrease of the maximum standardized uptake value (SUV_max) of at least 70% as measured by positron emission tomography (PET) using 2-[¹⁸F]fluoro-2-deoxy-D-glucose ([¹⁸F]FDG). According to Choi et al. [33, 34], a PR is defined as a ≥10% decrease in one-dimensional tumor size or a ≥15% decrease in tumor attenuation on contrast-enhanced CT scan, while PD is defined as a ≥10% increase in size without meeting PR criteria by change in attenuation (Table 1). Lesions with a longest diameter ≥15 mm are eligible for measurements according to the Choi criteria. In imatinib-treated patients with GIST, the Choi criteria had a significantly better correlation with disease-specific survival than RECIST.

Recently, we have evaluated the Choi criteria in 55 mRCC patients who were treated with sunitinib [22]. During sunitinib treatment, the median tumor attenuation decreased from 66 to 47 HUs. According to the Choi criteria, 36 (65%) patients achieved a PR, 6 (11%) patients had SD and 13 (24%) patients had PD at first evaluation. In 19 out of 36 responders, the achieved PR was only based on a decrease in tumor attenuation ≥15%. Patients with a PR according to Choi criteria had a prolonged PFS and OS as compared to patients without a PR. At first evaluation, the Choi criteria had a significantly better predictive value for PFS and OS than RECIST. However, the predictive value of the Choi criteria was similar to that of RECIST at later time-points. The Choi criteria were not able to early identify patients with PD. Therefore, we concluded that the Choi criteria will not change the clinical management of mRCC patients treated with sunitinib.

The Choi criteria may be more useful to identify mRCC patients with PD during sorafenib treatment, as sorafenib is associated with extensive necrosis [35] and fewer patients have tumor response to sorafenib as compared to sunitinib [32]. Two studies have described less favorable results for the use of the Choi criteria in comparison with RECIST in mRCC patients [32, 36]. In these analyses, however, fewer mRCC patients were included [36], different TKIs were administered (sunitinib, sorafenib, and cediranib) [32, 36], lung lesions were excluded for attenuation measurements [32], and attenuation measurements were based on volumetric data instead of two-dimensional data acquired in most representative axial images [32].

Nathan et al. [36] have compared the Choi criteria with the modified Choi criteria in which changes in both size and attenuation of target lesions have to be calculated to define an objective response (Table 1). In that study, CT scans from 32 mRCC patients who were treated with either sunitinib or cediranib were evaluated. Ten patients needed to be excluded from further analysis because of non contrast-enhanced CT scans due to renal impairment. Ultimately, 20 patients were evaluable of whom 5, 19, and 13 patients achieved a PR according to, respectively, RECIST, the Choi criteria, and the modified Choi criteria. The group concluded that the modified Choi criteria provided the best segregation of the median time-to-progression.

As a ≥10% decrease in tumor size or a ≥15% decrease in tumor attenuation is not necessarily associated with a prolonged PFS, the size and attenuation CT (SACT) criteria have been developed (Table 1) [32]. According to these criteria a favorable response is defined as a decrease in tumor size of ≥20% or a decrease in mean attenuation of ≥40 HU in at least one non-lung target lesion. In case patients have a decrease in tumor size of ≥10% and a decrease in mean attenuation of ≥20 HU in half of the non-lung target lesions, they are also classified as favorable responders. Patients have an unfavorable response in case of an increase in tumor size of ≥20% or the development of new lesions. Patients also have an unfavorable response in case of new enhancement in a homogeneously hypo-attenuating non-enhancing lesion or in case of a marked change from central necrosis to near complete enhancement of solid parts in the central area of the tumor. According to the SACT criteria, mean attenuation measurements are based on volumetric measurements. Lung lesions are not eligible because of inconsistent mean attenuation results from averaging between soft tissue and air. This is a potential limitation for the use of SACT criteria in mRCC patients as the lung is a common metastatic site. The use of absolute changes instead of percentage changes in tumor attenuation is a potential advantage of the SCAT criteria as compared to the Choi and the modified Choi criteria, since a 15% decrease in attenuation is reached early in lesions with low baseline values and may therefore result in less accurate measurements.

Although incorporation of changes in tumor attenuation appears to be valuable for response assessment, several limitations should be mentioned for its use in mRCC patients who are treated with targeted agents. A series of requirements have to be taken into account, such as identical scanning protocols and the same timing of intravenous contrast (IV) between the subsequent CT scans, as slightly different phases in IV contrast may result in incorrect changes in lesion attenuation. It should also be considered that targeted agents may affect the cardiac output [37], because of which the distribution of IV contrast may change during treatment despite the use of an identical scanning protocol. Additionally, contrast-enhanced CT scanning may not be possible in a number of mRCC patients, since administration of IV contrast may be contraindicated in case of an impaired renal function [36].

DCE-MRI is widely used in the diagnosis and staging of cancer and is emerging as a promising method for monitoring tumor response to targeted therapy [43, 44]. Quantitative T₁-weighted images of the microvasculature are obtained by tracking the pharmacokinetics of an IV bolus of a gadolinium-containing contrast agent, which is freely diffusible between the intravascular and the extracellular extravascular space. Gadolinium ions are paramagnetic and shorten the T₁-relaxation times in tissue water by interaction with nearby hydrogen nuclei. In a T₁-weighted DCE-MRI, the tumor enhancement over time reflects the delivery of the gadolinium contrast into the tumor tissue. Although DCE-MRI imaging protocols differ, three types of imaging data are usually acquired. The first images are used for localization of the tumor. Thereafter, sequences are acquired that are used to calculate the baseline T₁-values which are necessary for the analysis. At last, dynamic imaging data are acquired for about 5–10 min after gadolinium injection [44].

Data analysis of DCE-MRI is complex, as the obtained variables are mainly dependent on the acquisition protocol and the method for data analysis [45]. In addition, no consensus has been reached on the kinetic model to be used for the analysis of DCE-MRI data to measure the efficacy of targeted agents. The various acquisition and analysis methods have considerable influence on the interpretation of the obtained variables and their value as surrogate end-points for targeted therapy. For pharmacokinetic modeling of contrast agent uptake, a compartment model is commonly used to estimate three physiological parameters including the volume transfer constant K ^trans between the blood plasma and extravascular extracellular space (EES) (min⁻¹), the volume of EES per unit volume of tissue and the flux rate constant between EES and plasma (min⁻¹) (Fig. 2). A change in K ^trans of >40% is usually considered a true difference caused by drug effect [45]. In addition, the area under the gadolinium concentration curve is another frequently used parameter to determine treatment effects.

In a number of clinical studies in solid tumors, DCE-MRI has been applied to monitor early response to targeted therapy [46‐48], particularly in mRCC [49‐52]. In a phase II trial in 17 mRCC patients, sorafenib treatment induced a large decline in K ^trans of 60.3% [49]. The percentage decline of K ^trans and the change in tumor size by CT scan were significantly associated with PFS [49]. Additionally, high tumor K ^trans before treatment was also significantly associated with improved PFS [49]. In another DCE-MRI study in 44 mRCC patients, four weeks after initiation of placebo, sorafenib 200 mg × 2 and sorafenib 400 mg × 2, the mean K ^trans log ratios had changed by, respectively, +14%, −14% and −24% [50]. The changes in K ^trans were not associated with PFS, but patients with a high K ^trans at baseline had a prolonged PFS. The association between a high K ^trans at baseline and a prolonged PFS is in line with the results of the study by Flaherty et al. [49]. Remarkably, the effects of sorafenib on DCE-MRI parameters appeared to be highly variable across the tumor regions of interest which suggests that a more detailed voxel-based analysis, such as parametric images, might be more useful to analyze sorafenib-treated RCC lesions.

ASL is an alternative for the measurement of tumor blood flow by MRI and requires no external tracers [53, 54]. Labeling of arterial blood is achieved by spatially selective inversion of the MRI signal of inflowing arterial blood. The MRI signal from inverted blood is made negative relative to uninverted blood. When the labeled blood with a negative signal enters the tumor lesion, it attenuates the signal from the image of that lesion. The difference between the labeled image and the control image without labeling gives a measure of blood flow. The ASL tracer is water and can consequently be considered as highly diffusible.

In 10 mRCC patients treated with vatalanib, ASL MRI already showed changes in tumor blood flow at 1 month on treatment, while changes in size according to RECIST were only measured at 4 months [55]. ASL MRI demonstrated a significant decrease in blood flow (−42 ± 22%) in mRCC patients with SD or PR, whereas patients with PD had a non-significant increase in tumor blood flow (+25 ± 3%). As ASL MRI does not involve administration of contrast medium, the technique provides the opportunity of frequently repeated measurements, particularly in RCC patients with impaired renal function [56], and can be applied to determine the optimal time-point for perfusion measurement during antiangiogenic treatment.

DW-MRI is another alternative to contrast-enhanced MRI, as this technique does not require administration of contrast [57]. DW-MRI is sensitive to the random motion of water molecules and allows for non-invasive characterization of biological tissues on the basis of their water-diffusion properties [58]. Highly cellular tissues can be differentiated from acellular tissues, as diffusion of water molecules is reduced in the cellular area in comparison with that in the extracellular space. The degree of water diffusion within tissues, or the apparent diffusion coefficient (ADC) in mm²·s⁻¹, can be calculated by acquiring multiple images with different amplitudes or durations for the diffusion-weighting gradients. Consequently, increased cellularity and decreased interstitial space result in restriction of water molecules and decreased ADC values. In tumors, DW-MRI can be used to differentiate highly cellular from acellular regions and to monitor changes in cellularity over time, which reflects tumor response to therapy [59‐63]. In addition, ADC is a reproducible physical constant, which is independent of scanner, magnetic field and operator. However, there are some limitations with regard to the placement of regions of interest (ROI) on ADC maps caused by poor resolution, which results in partial volume effects.

In RCC, DW-MRI has been mainly evaluated for characterization of primary tumors [57, 64‐68]. In these studies, ADC values of high-grade clear cell carcinomas were lower than those of low-grade clear cell carcinomas [66], whereas ADC values of non-clear cell carcinomas were higher than those of the clear cell subtype [64]. The experience with DW-MRI in monitoring of mRCC during treatment is limited [69] and studies are warranted to substantiate the use of DW-MRI as a biomarker for response to targeted therapy in mRCC.

[¹⁸F]FDG PET is an important functional imaging tool in oncology [81]. The use of [¹⁸F]FDG is based on the fact that malignant tumor cells have a high glucose metabolism and actively take up glucose and its analogue [¹⁸F]FDG. The use of [¹⁸F]FDG PET in patients with RCC is limited [82], since [¹⁸F]FDG   PET has no major value for the detection of renal masses [83] or RCC metastases [84, 85]. As RCC has only a low potential for metabolizing glucose [86], a negative [¹⁸F]FDG PET does not rule out the presence of RCC. A positive [¹⁸F]FDG PET in RCC patients, however, is predictive for the presence of a RCC lesion [87, 88] and may complement anatomic imaging modalities, such as CT or may even be used in combination with CT [89].

When RCC is [¹⁸F]FDG avid, [¹⁸F]FDG PET can be applied to monitor the disease and may have additional value for response evaluation of targeted therapy [90‐93]. In mRCC, a significant decrease of [¹⁸F]FDG uptake has been measured in patients treated with either sunitinib or sorafenib [92, 93]. In a recent study, 12 mRCC patients underwent [¹⁸F]FDG PET to assess tumor response to sunitinib [93]. After the first treatment cycle, there was a decrease in the maximum standardized uptake value (SUV_max) of 17% (range; −59% to +100%), while the sum of the target lesions remained unchanged [93]. In two out of 12 patients, a non-significant decrease in SUV_max was measured, along with a non-significant increase in tumor size which was associated with a decrease in tumor density on CT [93]. As both patients had a long PFS (335 and 455 days), these findings suggested that the increase in tumor size in these patients was associated with the development of necrosis instead of PD. In another study, combined [¹⁸F]FDG PET/CT has been applied for assessment of tumor response in 10 mRCC patients treated with sorafenib [92]. After 1–2 months of sorafenib treatment, the mean and maximum [¹⁸F]FDG uptake in all lesions decreased to, respectively, 75% (range; 32–105%) and 86% (range; 46–131%) of the baseline value [92]. Sorafenib appeared to induce a heterogeneous decrease in [¹⁸F]FDG uptake in tumor lesions [92], suggesting that volumetric measurements may have additional value to determine tumor response in individual lesions. Both studies described above had performed [¹⁸F]FDG PET scanning after the first treatment cycle, but this technique may already be able to detect a metabolic response within 1 week as has been demonstrated for imatinib-treated GIST lesions [94, 95]. Indeed, an early metabolic response by [¹⁸F]FDG PET was reported within 2 weeks of therapy in a mRCC patient treated with semaxanib (SU5416) [90]. [¹⁸F]FDG PET has not yet been evaluated in mRCC patients on treatment with mTOR inhibitors, but early inhibition of [¹⁸F]FDG uptake has been noted in patients with non-small cell lung cancer who were treated with everolimus [96]. This suggests that [¹⁸F]FDG PET may also be useful to assess metabolic response in mRCC patients treated with mTOR inhibitors.

Although the role of [¹⁸F]FDG PET for response evaluation of targeted therapy is expanding [92, 93, 96, 97], [¹⁸F]FDG PET scanning is not incorporated in commonly used response evaluation criteria. According to RECIST, however, the use of [¹⁸F]FDG PET imaging is accepted as an adjunct to determination of progression [23]. The PET Criteria in Solid Tumors (PERCIST) version 1.0 has been proposed for metabolic tumor response assessment and serves as a starting point for PET-based response evaluation in clinical trials [98]. To ensure that PET images are acquired and interpreted optimally, [¹⁸F]FDG PET scans should be performed according to guidelines that provide a minimum standard for acquisition and require a common quality control and quality assurance [99].

Besides the most commonly used PET tracer [¹⁸F]FDG, other PET tracers have been developed for use in patients with malignancies. The accumulation of the PET tracer [¹¹C]acetate in tumor cells is related to the highly active lipid metabolism in the cell membrane associated with tumor growth. [¹¹C]acetate is channeled into the tricarboxylic acid cycle via acetyl coenzyme A and then incorporated into the cell membrane’s phospholipids. In malignancies, high uptake of [¹¹C]acetate was first reported in RCC [100, 101]. Currently, [¹¹C]acetate PET has been introduced as a new modality for imaging prostate cancer and its metastases [102]. In a RCC case report, uptake of [¹¹C]acetate in a non-[¹⁸F]FDG-avid hepatic metastasis has been described, which showed a reduction in [¹¹C]acetate uptake after only 2 weeks of sunitinib treatment [103].

With regard to the high vascularization in RCC lesions, quantification of tumor perfusion using radioactive water (H ₂ ¹⁵ O) may be used to monitor antiangiogenic treatment [91, 104‐106]. The injected radioactive water is freely diffusible in tissue [107] and a time-activity curve (TAC) reflecting ¹⁵O accumulation in tumors can be generated [108, 109]. The short half-life of radioactive oxygen (~2 min) enables serial measurements within a single scan session, but requires the presence of a nearby cyclotron for probe preparation.

Proliferation is an important feature of tumors and can be visualized using 3-deoxy-3-[¹⁸F]fluorothymidine ([¹⁸F]FLT) [110]. In patients with recurrent gliomas treated with bevacizumab and irinotecan, [¹⁸F]FLT PET seemed to be an early marker of OS, which was already predictive at 1–2 weeks after the start of treatment [111]. Although [¹⁸F]FLT uptake was measured in renal transitional cell carcinoma [112], it is currently unknown whether [¹⁸F]FLT uptake is a common characteristic of RCC.

As the PET technique is able to quantify tracer concentrations in absolute units, radiolabeling of targeted agents with short-lived positron-emitting radionuclides, such as carbon-11 [¹¹C] (half-life: 20.4 min) and fluorine-18 [¹⁸F] (half-life: 109.8 min), is an attractive approach to assess their uptake in tumor lesions. In dynamic PET studies, the optimal kinetic model of an injected tracer can be developed to quantify the tumor uptake of the radiolabeled drug [113]. Currently, several targeted drugs have been radiolabeled as PET tracers and their number is still increasing [114]. These specific PET tracers provide a unique opportunity for personalized treatment planning, since they may be able to predict tumor response before initiation of therapy [114]. Of the targeted agents for treatment of mRCC, [⁸⁹Zr]bevacizumab and [¹⁸F]sunitinib have been developed as PET tracers [115, 116]. PET studies with [⁸⁹Zr]bevacizumab [117] and [¹⁸F]sunitinib may not only be used to explore a potential relationship between pre-treatment uptake and response, but might also give insight into drug uptake during treatment and the development of tumor resistance.

Up to 30% of patients with mRCC present with metastatic disease. In the era of immunotherapy, response of the primary tumor to cytokine-based therapy was very rare [118‐120]. Therefore, patients with mRCC and a resectable primary tumor in situ usually underwent cytoreductive nephrectomy. This treatment strategy was based on the results from two randomized phase III clinical trials demonstrating a longer OS in patients with nephrectomy followed by interferon-α versus patients on interferon-α alone [121‐123]. In the era of targeted therapy, however, cytoreductive nephrectomy in patients with asymptomatic primary tumors has become controversial [124], as impressive drug-induced responses have been observed in primary RCC tumors [21]. Moreover, targeted therapy can palliate primary tumor-related symptoms, such as pain and hematuria.

The observed responses in primary RCC lesions have provided opportunities for new treatment approaches in patients who present with synchronous metastases. First, treatment with targeted agents may make cytoreductive nephrectomy unnecessary, preventing patients from peri-operative and post-operative complications. A prospective randomized phase III clinical trial is underway addressing the question whether refraining patients from nephrectomy has an impact on their survival. Second, targeted agents can be applied as neoadjuvant therapy to achieve downsizing of the primary tumor that may facilitate cytoreductive nephrectomy [20, 125‐127]. A few treatment cycles may be sufficient, as the optimal effect of downsizing primary tumors is mostly achieved in the first 3 months [21]. The PD rate in primary tumors, however, varies from 0% to 47% [21, 128] and mixed responses between the primary tumor and the metastases may occur, indicating that there is a risk of progression of initially resectable primary tumors to inoperability despite a response in metastases [129]. Therefore, appropriate response evaluation is required to early identify mRCC patients with failure of targeted therapy in the primary tumor. Then, when patients develop PD in resectable primaries in the presence of stable disease in metastases, the treatment plan can be changed rapidly and cytoreductive surgery can be reconsidered.

When using RECIST to determine treatment response in patients with both a primary tumor and metastases, the overall response in these patients may be underestimated. In general, primary tumors of RCC have an enormous size as compared to their metastases [130, 131]. In addition, large primary tumors show a relatively smaller decrease in size than their metastases during targeted therapy [21, 132]. Consequently, inclusion of large primaries in RECIST measurements can have significant impact on the overall objective response. For that reason, it may be preferable to determine the response in primary tumors and metastases separately and exclude primaries from overall response assessment [133].

Volumetric measurements may give more insight into the response of primary tumors during targeted therapy. We did not find volumetric measurements of sunitinib-treated primary tumors of additional value [21], because one-dimensional changes in size showed a high correlation with volumetric changes. This study indicated that changes in size of the primary tumors can be considered symmetrical and that RECIST measurements are sufficient for evaluation of response to sunitinib. Note that for changes in tumor size, the response criteria are different for one-dimensional (RECIST), two-dimensional (World Health Organization definitions) and three-dimensional (volumetric) measurements, and a PR is reached upon a decrease in tumor size of, respectively, 30%, 50% and 65% [23].

As targeted therapy in primary tumors is associated with extensive tumor necrosis, assessment of changes in tumor attenuation may be an additional indicator of response. Targeted therapy consisting of sunitinib induced a 31% reduction in the median volume of the primary tumor and was accompanied by a 39% increase in the median volume of primary tumor necrosis [21]. In addition, the median decrease in enhancement as measured in one plane of the primary tumor was 13% in primary tumors treated with sorafenib [126]. With regard to changes in tumor attenuation in primary tumors, the largest series has been described by Cowey et al. [134]. In this retrospective study of 30 mRCC patients treated with sunitinib or sorafenib, a positive relationship between the loss of primary tumor enhancement and the degree of primary tumor shrinkage was seen: a larger percentual decrease in treatment-induced enhancement correlated with a larger percentual decrease in primary tumor size. No relationship was found between the pre-treatment contrast-enhancement and the pre-treatment size of the primary tumor. There were a few patients in whom the change in dimension did not match the degree or direction of change in enhancement, but the clinical significance of this discrepancy is not clear.

Lung metastases can be detected in up to 80% of mRCC patients [135]. Lung lesions can have a rather small size, which may limit specific response measurements. As an example, we have reported that a large number of lung metastases was ineligible for measurement according to the Choi criteria, as these lesions had a size <15 mm at baseline [22]. These lesions were still eligible for measurements according to RECIST, as their size was ≥10 mm. In addition, assessment of tumor attenuation in lung lesions implies a risk of unreliable mean values due to averaging of attenuation values between soft issue and air.

Targeted therapy is known to cause air-containing cavitations in lung lesions of several solid tumors, including RCC [22] (Fig. 5). Cavitations of lung lesions are especially common in patients with non-small cell lung cancer on treatment with inhibitors of VEGF signaling and platinum-based chemotherapy [136, 137]. Crabb et al. [137] have reported that response evaluation of lung lesions may be improved by incorporation of cavitations into the volume assessment. Incorporation of cavitations into response evaluation of lung metastases of mRCC is not required as the incidence of cavitations appears to be relatively low [22, 32], likely due to their relatively small size.

In RCC, the bone is the second most common site of metastases [135]. In patients without bone metastases, progression of the disease with respect to the bone is easily defined by the development of any new bone lesions during treatment. In case of pre-existent bone metastases, however, response evaluation of targeted therapy is more difficult [138]. Traditionally, all bone lesions were considered as non-target lesions according to RECIST version 1.0 [23]. RECIST version 1.1 states that osteolytic bone lesions with an identifiable soft tissue component, assessed by CT or MRI, can be considered as measurable if the soft tissue component otherwise meets the definition of measurability [23]. Osteoblastic bone lesions, however, are still truly non-measurable. Furthermore, PET scan, bone scan, or plain films are not considered adequate imaging techniques to evaluate bone lesions, but these modalities can be used to confirm the presence or disappearance of bone lesions [23, 139, 140]. Although several investigators have attempted to validate new criteria for evaluation of bone metastases [140], no consensus has yet been reached.

Currently, contrast-enhanced MRI is the best diagnostic test to detect brain metastases [141]. However, in patients with mRCC brain imaging is usually performed when patients present with central nervous system symptoms. Therefore, the experience on evaluation of targeted therapy in brain metastases from mRCC is mostly limited to symptomatic brain metastases. Presently, the reports on the efficacy of targeted therapy in brain metastases of RCC have been conflicting. Brain metastases have been described as the first sign of progression in mRCC patients treated with targeted agents suggesting relatively poor penetrance into the brain [142], whereas responses of brain metastases to targeted therapy have also been described [143‐146].

Response evaluation of targeted therapy has been mostly performed in brain metastases that have been pretreated with radiotherapy or surgery. This may make the interpretation of brain imaging during targeted therapy more difficult, as both radiotherapy and surgery cause radiographic changes. After surgery or radiotherapy of the brain, treatment with sorafenib and sunitinib in mRCC patients may cause an increase in peritumoral edema and MRI contrast-enhancement [147]. Such radiographic changes may result from leakage of the drug across the blood brain barrier that is already impaired by previous surgical or radiation intervention [147]. Misinterpretation of these radiographic changes may suggest PD and the applied treatment as inadequate. In such cases, however, discontinuation of the targeted drug may lead to progressive improvement of edema and MRI enhancement [147].

Brain metastases are associated with vasogenic cerebral edema, which is caused by leakage of fluid from the intravascular space into the brain parenchyma due to the hyperpermeability of the tumor vasculature. This abnormally permeable tumor vasculature is stimulated by VEGF that is secreted by tumor cells. Therefore, targeted agents that block the VEGF signaling pathway may be able to reduce tumor-associated edema by decreasing vascular permeability. In glioblastoma patients, cediranib, a TKI of VEGFR, caused rapid and significant reduction in vasogenic edema which was associated with a decrease in vascular permeability, as measured by the transfer constant K ^trans using DCE-MRI [148]. Although DCE-MRI has not yet been evaluated to assess the effects of targeted agents in brain metastases from RCC, similar effects may be expected with agents inhibiting VEGF signaling.