The role of early 18F-FDG PET/CT in prediction of progression-free survival after 90Y radioembolization: comparison with RECIST and tumour density criteria
- First Online:
- Cite this article as:
- Zerizer, I., Al-Nahhas, A., Towey, D. et al. Eur J Nucl Med Mol Imaging (2012) 39: 1391. doi:10.1007/s00259-012-2149-1
- 511 Views
This study evaluated the ability of 18F-FDG PET/CT imaging to predict early response to 90Y-radioembolization in comparison with contrast-enhanced CT (CECT) using RECIST and lesion density (Choi) criteria. Progression-free survival (PFS) in patients with liver metastases at 2 years and decline in tumour markers were the primary end-points of the study.
A total of 121 liver lesions were evaluated in 25 patients (14 men, 11 women) with liver-dominant metastatic colorectal cancer who underwent 18F-FDG PET/CT and CECT before and 6–8 weeks after treatment. Changes in SUVmax, tumour density measured in terms of Hounsfield units and the sum of the longest diameters (LD) were calculated for the target liver lesions in each patient. The patient responses to treatment were categorized using EORTC PET criteria, tumour density criteria (Hounsfield units) and RECIST, and were correlated with the responses of tumour markers and 2-year PFS using Kaplan-Meier plots and the log-rank test for comparison. Multivariate proportional hazards (Cox) regression analysis was performed to assess the effect of relevant prognostic factors on PFS.
Using 18F-FDG PET/CT response criteria, 15 patients had a partial response (PR) and 10 patients had stable disease (SD), while using RECIST only 2 patients had a PR and 23 had SD. Two patients had a PR, 21 SD and 2 progressive disease using tumour density criteria. The mean changes in SUVmax, sum of the LDs and tumour density after treatment were 2.9 ± 2.6, 7.3 ± 14.4 mm and 1.9 ± 13.18 HU, respectively. Patients who had a PR on 18F-FDG PET/CT had a mean decrease of 44.5 % in SUVmax compared to those with SD who had a decrease of only 10.3 %. The decreases in SUVmax and sum of the LDs were significant (p < 0.0001, p < 0.05, respectively) while the decrease in tumour density was not (p > 0.1065). The responses on the 18F-FDG PET/CT studies were highly correlated with the responses of tumour markers (p < 0.0001 for LDH, p = 0.01 for CEA and p = 0.02 for Ca19-9), while the responses on the CECT studies using both RECIST and tumour density criteria were not significantly correlated with the responses of tumour markers. The responses on 18F-FDG PET/CT studies also significantly predicted PFS (the median PFS in those with a PR was 12.0 months and in those with SD was 5 months, p < 0.0001), while RECIST and tumour density did not significantly predict PFS. Multivariate analysis demonstrated that responses on 18F-FDG PET/CT studies and decreases in SUVmax of ≤2.0 were the strongest predictors of PFS.
Early response assessment to 90Y-radioembolization using 18F-FDG PET/CT is superior to RECIST and tumour density, demonstrating a correlation with tumour markers and significantly predicting PFS in patients with liver metastases. This could enable early response-adapted treatment strategies to be employed.
Colorectal cancer patients presenting with liver metastases have poor overall survival, and surgical resection remains the only curative treatment option. However, only 20 % are deemed candidates for surgical resection and only 20 % of those who have undergone tumour excision will survive for 5 years . Nonsurgical treatment options have evolved in the past 20 years, and of these internal radioembolization has gained increasing popularity in the last decade. This therapy involves the administration of glass or resin 90Y-labelled microspheres directly into the hepatic arteries. Liver tumours are known to receive the majority of their blood supply via the hepatic arteries rather than the portal system  and therefore the injected microspheres will be selectively directed to the tumours. Once trapped in the microvasculature of the liver tumours, the microspheres deliver a high dose of ionizing radiation in excess of 100 Gy  with relative sparing of surrounding liver tissue.
Response assessment, using the Response Evaluation Criteria In Solid Tumours (RECIST) has become the gold standard in evaluating response in routine oncology practice and for assessing the efficacy of any novel cancer therapy . However, it has become widely recognized that these criteria, which rely solely on changes in tumour size, have limitations particularly in cytostatic treatments and therapies that produce tumour necrosis leading to an apparent increase in the size of the lesion. This is also the case when assessing response to 90Y-radioembolization where these criteria fail to depict the changes in treated liver lesions. In addition, changes in tumour size can take several months before becoming apparent which makes these criteria insensitive in early prediction of response to treatment. This is highly important in patients where the course of treatment will be altered as a result of response on imaging.
Alternative criteria such as measurement of viable tumour diameter on contrast-enhanced CT (CECT) after vascular interventional therapies of hepatocellular carcinoma (European Association for the Study of the Liver criteria) and evaluation of tumour density in terms of Hounsfield units (criteria of Choi et al.) have been proposed as alternatives to RECIST [5, 6]. Biomarkers of tumour metabolism have also been suggested as a more accurate way of assessing response to therapy [7, 8]. Of these, PET imaging with 18F-FDG has become an established imaging tool in the assessment of response to anticancer therapies in a variety of solid tumours as well as a prognostic indicator [8, 9]. Furthermore,18F-FDG PET/CT has been advocated in the assessment of response to cytostatic therapies such as the tyrosine kinase inhibitor imatinib for gastrointestinal stromal tumours and radiofrequency ablation of liver tumours where it has provided additional information on response to therapy [10, 11]. Standardized PET/CT response criteria have been proposed which include the European Organization for Research and Treatment of Cancer (EORTC PET criteria) [12, 13] and the most recently published PET response criteria (PERCIST) .
Response assessment post 90Y-radioembolization using 18F-FDG PET/CT in comparison with tumour density criteria and RECIST has not been previously studied. The aim of our study is to evaluate the role of early 18F-FDG PET/CT compared to RECIST and tumour density in assessing response to 90Y-radioembolization using tumour markers and for predicting 2-year liver progression-free survival (PFS) as gold standards.
Materials and methods
The study was compliant with departmental ethics policies and did not require research and development approval. The study group for this retrospective study comprised 25 patients selected between October 2008 and January 2010. Patients with dominant colorectal liver metastases and no more than one lung or one bone metastasis amenable to local treatment were included. Patients had standardized CECT and 18F-FDG PET/CT before and after radioembolization following our standard departmental imaging protocol for patients undergoing 90Y-radioembolization. All patients also underwent clinical examination, blood tests including full blood count, liver and renal function tests, lactate dehydrogenase (LDH), carcinoembryonic antigen (CEA) and carbohydrate antigen 19-9 (Ca19-9)). Patients were selected for 90Y-radioembolization at the multidisciplinary team (MDT) meeting (the inclusion criteria for treatment were age over 18 years, unresectable hepatic metastatic colorectal cancer, liver-dominant tumour burden, Karnofsky performance score ≥70 and a life expectancy of at least 3 months). Patients were considered not suitable for treatment if they had a history of external beam radiotherapy, ascites or severely abnormal liver function tests suggesting clinical liver failure.
Follow-up was performed in the oncology/hepatobiliary clinic 6 weeks after treatment, then 3-monthly for 2 years and every 6 months thereafter.
All patients underwent an initial selective angiographic evaluation of the abdominal and hepatic arteries with embolization of aberrant vessels, 1–2 weeks prior to the therapy. The coeliac axis was catheterized using a 7F Sidewinder catheter (Cordis, Johnson and Johnson Medical, Waterloo, Belgium) via a transfemoral approach and if required microcatheters (3F coaxial catheter and wire system, Tracker 325, Target Therapeutics, Fremont, CA; Renegade, Boston Scientific, Natick, MA; or Progreat Terumo, Somerset, NJ) were used to selectively catheterize the hepatic arteries.
An activity of 100 MBq 99mTc MAA was administered into the hepatic artery and the patient’s liver and lung were imaged in order to calculate the hepatic shunt to lungs. Planar images were obtained with a dual head gamma camera (E.cam; Siemens Medical Solutions, Malvern, PA) and low-energy high-resolution collimators. After 1–2 weeks, microspheres were administered (SIR-Spheres; Sirtex, Sydney, Australia) via an intrahepatic microcatheter under fluoroscopy guidance. The dose of the microspheres to be administered in each patient was calculated based on the body surface area (BSA) method adjusted for the percentage of 99mTc-MAA shunt. The injected activity was calculated using the equation: activity in gigabecquerels = (BSA in square metres − 2) + (percent tumour involvement/100)
All patients had an overnight hospital stay and were discharged the next day following satisfactory review.
All patients had an independent CECT scan 3–4 weeks before and 6–8 weeks after therapy (GE LightSpeed four-slice helical CT scanner; Waukesha, WI). Arterial (20 s) and portal venous (60 s) CT studies of the abdomen and pelvis (5-mm collimator, pitch 0.8, 120 kVp, 80 mAs) following intravenous injection of 100 ml of Omnipaque 300 were performed (GE Healthcare). Images were reconstructed to 3-mm axial slices using ordered-subsets expectation maximization (OSEM) iterative reconstruction and archived in the PACS (GE Healthcare) workstation. The latter was used to analyse the CECT images.
18F-FDG PET/CT imaging
All patients had a 18F-FDG PET/CT scan 3–4 weeks before treatment and 6–8 weeks after treatment using the same departmental protocol. PET/CT scans (Siemens Biograph 64-slice; Knoxville, T/Muenchen, Germany) were performed from the base of the skull to the upper thighs following a 6-hour fast. After injection of 370 MBq18F-FDG, emission data were acquired (five or six bed positions, 4 min per bed position) after a 60-min uptake period. The CT parameters were 120 KVp, 50 mAs, pitch 0.8, 5-mm slices with 3-mm separation. The PET data were reconstructed using OSEM iterative reconstruction and were attenuation-corrected using the CT data. The study was archived and displayed on the Siemens workstation for analysis.
The 18F-FDG PET/CT and CECT data were analysed by a single experienced dual accredited nuclear medicine radiologist at the same setting for each patient, but independently to avoid lesion measurement bias.
Selection of target lesions
Target lesions in the liver were defined as lesions with the longest diameters (LD) as defined by RECIST (version 1.1)  and the most metabolically active lesions on 18F-FDG PET/CT reflecting viable tumour burden. Lesions smaller than 1.5 cm or those which had 18F-FDG uptake less than twice the normal liver background uptake were excluded. The same lesions were used for both PET/CT and CECT as well as the imaging studies before and after treatment.
On the PET/CT images liver lesions with the most intense 18F-FDG uptake among all foci were identified in each patient. These were lesions with the highest SUVmax and had uptake at least twice the normal liver background uptake.
The SUVmax (normalized for BSA) was obtained for each target lesion on PET/CT studies both before and after 90Y-radioembolization using the Siemens PET/CT workstation and region of interest (ROI) analysis. The absolute and percent change in SUVmax was then calculated in each patient.
The LD of the target lesions was measured using the electronic callipers on the PACS workstation. Based on RECIST 1.1, the sum of the LDs in each patient was computed on the CECT studies before and after treatment. The absolute and percent changes in the sum of the LDs after 90Y-radioembolization was then calculated in each patient.
A ROI was also drawn around the target lesion to determine its density in terms of the mean Hounsfield units on the portal venous phase of the CECT studies based on the criteria of Choi et al. The tumour density measurements of all lesions were combined and a mean value for each patient was computed on the CECT studies before and after treatment. Again, the absolute and percent change in tumour density after 90Y-radioembolization was calculated.
EORTC PET criteria, RECIST and tumour density criteria (HU)
EORTC PET criteria
Tumour density criteria
Complete resolution of FDG uptake in lesions
Disappearance of measurable disease
Disappearance of all enhancing lesions
Reduction in SUV >25 %
≥30 % decline in sum of the LDs
Decrease in tumour density ≥15 %
No PD or PR/CR
No PD or PR/CR
No PD or PR/CR
New FDG uptake in metastatic lesion; increase in SUV >25 %; visible increase in extent of FDG uptake (20 % in LD)
New lesions; increase ≥20 % in the sum of the LDs and absolute increase of ≥5 mm
New lesions; increase ≥20 % in tumour density
The responses determined using tumour markers (CEA, Ca19-9 and LDH) following 90Y-radioembolization were compared with response categories on PET/CT and CECT.
PFS, the primary end-point of the study, was calculated from the date of treatment until the date of progression or relapse in the liver, or disease-related death. Patients were followed up in the oncology clinic at regular intervals to determine PFS (at 6 weeks after treatment then 3-monthly for 2 years) and all data (clinical status, blood results, tumour markers) were documented in the clinical notes. If there were any suspicious clinical signs of progression, imaging studies were also requested (CECT or PET/CT). Progression was defined on imaging using RECIST 1.1 for progression on CECT or EORTC criteria on 18F-FDG PET/CT.
Responses determined from CECT and PET/CT studies were correlated with responses in tumour markers using the Mann-Whitney test to determine statistical significance (p < 0.05 considered as significant). PFS in the cohort was depicted using Kaplan-Meier plots. Differences between response groups were analysed using the log-rank test. Data were censored if progression was outside the liver only, on death from other causes or if the patients were free of progression/relapse at the last follow-up. Multivariate proportional hazards (Cox) regression analysis was performed to assess the effect of relevant prognostic factors on PFS and the independence of these variables. The following parameters were used: changes in the sum of the LDs, tumour density, SUVmax, response on 18F-FDG PET/CT studies, clinical stage, extent of liver disease, extent of disease outside the liver, and subsequent treatments. All statistical analyses were performed using the IBM SPSS Statistics version 19.0 package (IBM corporation, Somers, NY).
Demographic information and patient follow-up
Number of patients
Percent of patients
Location of metastases
Liver and lung
Liver and bone
Primary tumour resected
Number of liver segments involved
1 or 2
2 to 4
4 or 5
5 to 8
Before therapy (>1 month)
From October 2008 to January 2010, 25 patients selected during the hepatobiliary MDT meeting were treated with 90Y-radioembolization as a salvage procedure. It was considered that five patients might benefit from chemotherapy, and these patients also received adjuvant 5-fluorouracil, leucovorin and oxaliplatin (FOLFOX). Two patients from the cohort had single pulmonary metastases which were amenable to local treatment such as radiofrequency ablation. One patient had a solitary bone lesion which was treated with radiotherapy. No complications were reported during or after 90Y-radioembolization.
Response assessment on imaging studies
SUVmax, sum of the LDs and tumour density values before and after treatment
Sum of LDs (mm)
Tumour density (HU)
After treatment, there were decreases in mean SUVmax (2.9 ± 2.6; 95 % CI 1.8, 3.9; 30.1 % decrease), the sum of the LDs (7.3 ± 14.4 mm; 95 % CI 1.7, 12.9 mm; 12.5 % decrease), and tumour density (1.9 ± 13.18 HU; 95 % CI −3.66, −7.47; 7.5 % decrease). The decreases in SUVmax and the sum the LDs were statistically significant (p < 0.0001 and p < 0.05, respectively), but the decrease in tumour density was not (p > 0.1065).
Response categories in relation to the evaluation criteria
EORTC PET criteria
Tumour density criteria
The 18F-FDG PET/CT criteria showed a significantly greater number of patients with a PR (n = 15) than the CECT based criteria (n = 2 each with RECIST and tumour density criteria). There was poor agreement between the response evaluations using the 18F-FDG PET/CT and CECT criteria (weighted kappa −0.027), which were identical in only 44 % of patients.
Correlation of response on 18F-FDG PET/CT and CECT with blood markers
All patients who achieved a PR on 18F-FDG PET/CT had a decrease in LDH (mean 67 %, median 75 %), 11 patients with a PR had a decrease in CEA (mean 74 %, median 78 %) and 9 patients had a decrease in Ca19-9 (mean 56 %, median 65 %). No significant decreases in tumour markers were observed in patients with SD by 18F-FDG PET/CT (mean decreases in CEA, LDH and Ca19-9: 9.7 %, 7.6 % and 8.6 %, respectively). There was a statistically significant correlation between the decreases in tumour markers and the responses on 18F-FDG PET/CT (percent LDH , CEA and Ca19-9 responses: p < 0.0001, p = 0.01 and p = 0.02, respectively, Mann-Whitney test). RECIST and the tumour density criteria failed to demonstrate any correlation between response and tumour markers (p = 0.265).
Two patients had extrahepatic progression with lung metastases, but their intrahepatic disease showed a PR. Two patients had a dramatic reduction in the volume of their liver disease after 90Y-radioembolization permitting subsequent surgical resection. One patient was lost to long-term follow-up. These five patients were censored.
The median PFS for the whole study population was 9 months. There was a significant longer PFS in patients who showed a PR on the early 18F-FDG PET/CT studies compared to those who showed SD (p < 0.0001). The median PFS was 12 months (95 % CI, 9.9–15.4 months) in patients with a PR and 5 months (95 % CI, 3.8–6.1 months) in those who had SD. Patients who had a PR had a mean decrease in SUVmax of 44.5 % (mean decrease 4.33, range 2.11–9.23) compared to those with SD who a decrease of only 10.3 % (mean decrease 0.99, range −0.75–2.06).
Using RECIST and tumour density criteria, there was no clear separation between patients who responded to 90Y-radioembolization and those who did not. The two patients with a PR using RECIST had a PFS of 9 and 4 months, and those with SD had a median PFS of 8 months. Similarly, the patients with SD, PD and PR using the tumour density criteria had a median PFS of 8, 9 and 14 months, respectively. No meaningful statistical analysis could be performed to assess the difference in these categories.
Results of the Cox multivariate analysis of the significance of various factors in predicting PFS
95 % CI
Clinical stage (stage IV)
Extent of liver disease (five to eight segments involved)
Extent of disease outside the liver
Locally advanced primary
PET/CT after treatment
No response (SD)
SUVmax change ≤2.0
SUVmaxon PET/CT before treatment
≥5 to <10
≥10 to <15
No change in tumour size
No change in tumour density
To evaluate the optimal cut-off value of percent SUVmax decrease that was able to predict PFS and an early response after 90Y-radioembolization, Kaplan-Meier plots were generated using percent SUVmax decreases of 5 %, 10 %, 15 %, 20 %, 25 %, 30 %, 35 % and 40 %, and the difference in PFS between patients with SD and those with PR was evaluated. Percent SUVmax values of 25–30 % correlated best with PFS.
Several previous studies have shown that18F-FDG PET is a more sensitive and accurate predictor of response to therapy using visual and/or semiquantitative analysis [16–18]. An early prospective study of eight patients treated with 90Y-labelled glass microspheres evaluated 18F-FDG PET compared to CT or MRI in response assessment 3 months after treatment . There was a significant metabolic response assessed visually on the 3-month follow-up 18F-FDG PET study, which correlated with a decrease in serum CEA. However, anatomic assessment on the CT and MRI studies did not reveal a significant change in the size of the liver lesions and showed no correlation with the decrease in serum CEA. Popperl et al. reported their early experience in the use of 90Y-radioembolization . They assessed the responses in 23 patients with nonresectable liver disease not responding to chemotherapy or local treatments by 18F-FDG PET. Marked decreases in 18F-FDG uptake of liver metastases were seen in 10 of 13 patients after 3 months, and in 3 of these 10 patients normal hepatic 18F-FDG uptake was achieved. These decreases paralleled decreases in tumour markers (CEA, Ca19-9 and CA15-3, and the special markers for neuroendocrine tumours Cyfra 21-1, ProGRP and NSE) in all 10 patients, and in 5 of these 10 patients normal levels were achieved approximately 3 months after 90Y-radioembolization. CT only demonstrated a slight decrease or SD in tumour load in the liver.
Our study assessed response in FDG PET studies at 6–8 weeks, rather than at 3 months as in the two previously mentioned studies, as 3 months was felt to be a reasonable interval to identify patients who might benefit from adjuvant treatment, but the optimum time point for response assessment is not known. A later study compared visual assessment to quantitative analysis of total liver uptake using SUV summed in four to eight resliced axial images of the liver. The mean percentage reduction in tumour metabolism measured in terms of total liver uptake was significantly higher in responders (−26 ± 15 %) than in nonresponders (+6 ± 15 %, p = 0.004) . Our study is not directly comparable as using total liver FDG uptake may oversimplify matters as to the assessment of individual lesions. In particular total liver uptake may not be sensitive in patients with multiple small foci or a relatively small amount of metabolically active tumour compared to the mass of the liver.
An early study performed at our institution that compared 18F-FDG PET and RECIST in 21 patients treated with radioembolization demonstrated that an early 18F-FDG PET study performed at 6 weeks showed a decrease in SUV in 86 % of patients, while CT demonstrated a decrease in tumour size in only 13 % of patients . The mean SUV before treatment was 12.2 ± 3.7 and after treatment was 9.3 ± 3.7 (p = 0.01). In one patient the decrease in 18F-FDG uptake allowed downstaging, and as a consequence the patient underwent surgical resection. In the current study two patients had a marked reduction in liver disease allowing subsequent liver resection.
Tumour density criteria which were proposed as alternative means for measuring response to molecular targeted and locoregional therapies in liver tumours were worth investigating in our study. To our knowledge, there are no studies in the literature that have compared changes in tumour density on CECT studies to changes in SUVmax on 18F-FDG PET/CT studies after 90Y-radioembolization. As the amount of viable tumour reduces and the amount of necrotic tissue increases after treatment, we expected a parallel decline in mean tumour density. However, to the contrary, the present study demonstrated that overall lesion enhancement did not show a significant decrease after 90Y-radioembolization. This was probably due to microinfarctions that occurred in the treated liver lesions leading to microhaemorrhages within the tumours resulting in an increase in tumour density. These changes can take up to 6 months to resolve, making tumour density criteria insensitive in predicting early response to treatment. Furthermore, the overall change in tumour density alone may be an oversimplification in the assessment of viable tumour. Evaluation of dynamic tumour enhancement is probably more accurate, but this is complex to perform and quantify on CT .
The current study is also the first of its kind to correlate early 18F-FDG PET/CT response with PFS in patients with liver metastases. Patients who had a PR had a median PFS of 12 months compared to 5 months in those with SD. 18F-FDG PET/CT response and change in SUVmax after 90Y-radioembolization were the strongest predictors of PFS in the multivariate analysis. Further analysis of PFS demonstrated that a 25–30 % decrease in SUVmax was the best predictor of PFS in this study group. The recently described PERCIST criteria include a 30 % reduction in SUV corrected for lean body mass based on reproducibility data and the lower likelihood that this level of decline occurs by chance . In our study, defining a PR as a 30 % reduction in SUVmax instead of a 25 % reduction would not have changed our results because none of our patients had a reduction between 25 % and 30 %. It is recognized that these criteria are to facilitate trials, as the optimum cut-off is treatment- and tumour-dependent. However, although this was a small retrospective study, the results are encouraging and justify further evaluation in a larger study to check reproducibility and the optimum cut-off value for response assessment.
This study demonstrated that early 18F-FDG PET/CT (metabolic response criteria) is superior to CECT (RECIST and tumour density) in predicting PFS in patients with liver metastases and tumour marker responses after 90Y-radioembolization.
Conflicts of interest