CLINICAL STUDY

High Lung Shunt Fraction in Colorectal LiverTumors Is Associated with DistantMetastasis and Decreased Survival

Amy R. Deipolyi, MD, PhD, A. John Iafrate, MD, PhD,Andrew X. Zhu, MD, PhD, Emel A. Ergul, MA,

Suvranu Ganguli, MD, and Rahmi Oklu, MD, PhD

ABSTRACT

Purpose: To assess how intratumoral shunting relates to liver metastasis and to clinical outcome.

Materials and Methods: Lung shunt fraction (LSF) was calculated from macroaggregated albumin scan after transcatheterinjection of radioactive particles in 62 patients with colorectal cancer and liver metastases evaluated for selective internalradiation therapy (SIRT) from May 2007 to August 2012. Assessment was performed of how LSF, liver tumor burden, andsystemic chemotherapy relate to survival and the presence of lung metastases. LSF and tumor burden were also assessed in asubset of patients who underwent genetic profiling with SNaPshot analysis.

Results: Patients with higher LSF were more likely to have lung metastases and decreased survival, whereas tumor burden wasnot associated with these outcomes. Patients with genetic mutations had significantly higher LSF than patients with nomutations. Patients who received chemotherapy before SIRT and had low LSF had the longest survival after SIRT.

Conclusions: LSF may be a more robust marker of metastasis than tumor size. Increased LSF secondary to vascular shuntingwithin liver metastasis is an indicator of distant lesions and is associated with decreased survival after SIRT. Intratumoralshunting may provide a conduit for circulating tumor cells to access more remote organs, bypassing filtration by liverparenchyma, and may be an important factor in metastasis from colorectal cancer.

ABBREVIATIONS

LSF = lung shunt fraction, SIRT = selective internal radiation therapy

The mechanisms underlying metastasis of colorectalcarcinoma and associated genetic mutations are poorlyunderstood (1–3). Primary tumors shed circulating tumorcells typically 20–30 μm in diameter (4). It is unclear howthese large cells traverse capillary beds 8–10 μm indiameter to establish distant lesions. One possibility isthat intratumoral vascular shunting provides a conduitfor circulating tumor cells to access distant sites (1).

& SIR, 2014

J Vasc Interv Radiol 2014; 25:1604–1608

http://dx.doi.org/10.1016/j.jvir.2014.06.019

From the SIR 2014 Annual Meeting.

None of the authors have identified a conflict of interest.

From the Divisions of Interventional Radiology (A.R.D., E.A.E., S.G., R.O.),Molecular Pathology (A.J.I.), and Hematology/Oncology, Center for LiverCancer (A.X.Z.), Massachusetts General Hospital, Harvard Medical School,55 Fruit Street, 290 Gray/Bigelow, Boston, MA 02114. Received April 21,2014; final revision received June 22, 2014; accepted June 23, 2014. Addresscorrespondence to R.O.; E-mail: roklu@mgh.harvard.edu

Arteriovenous shunt formation within tumors couldallow these cells to bypass capillaries. For example,circulating tumor cells emerging from colorectal cancercould form lung metastases by bypassing normal hepaticcapillary beds through shunts within hepatic metastases.Endovascular hepatic arterial procedures have shown

that intratumoral vascular shunting is common (5,6).Assessment of the degree of arteriovenous shunting inliver tumors is required before selective internal radia-tion therapy (SIRT) (7). After transarterial injection oftechnetium-99m-labeled macroaggregated albumin par-ticles measuring 30–150 μm in the left or right hepaticartery, the fraction of radioactivity detected in the lungsis calculated and expressed as the lung shunt fraction(LSF) (7). Representing the extent of intratumoralshunting within liver lesions, LSF determines theradiation dose that may be safely administered withoutcausing radiation pneumonitis.To determine whether intratumoral shunting is associ-

ated with metastasis, the relationship of tumor shunting

Volume 25 ’ Number 10 ’ October ’ 2014 1605

to the presence of disseminated disease and clinicaloutcome in colorectal cancer was assessed. Specifically,how LSF, compared with liver tumor burden, related tothe presence of distant metastases in the lungs and tooverall survival was characterized. The association ofgenetic mutations with shunting and metastasis was ana-lyzed. Additionally, how the use of chemotherapy impa-cted the degree of intratumoral shunting was studied.

MATERIALS AND METHODS

All patients evaluated for SIRT (May 2007–August2012) were included in this institutional review board–approved retrospective study (Table). Work-up beforeSIRT included computed tomography scans of the chest,abdomen, and pelvis followed by evaluation in interven-tional radiology, as previously described (7). Briefly, thisevaluation included hepatic angiography, embolizationof the gastroduodenal and other arterial branches toprevent nontarget radioembolization, and transcatheterinjection of radioactive particles (technetium-99m-labeled macroaggregated albumin) into the artery plan-ned for treatment, usually the right or left lobar hepaticartery. Planar and single photon emission computedtomography imaging of the head, chest, and abdomen

Table. Summary of Patient Demographics

Age (y) 62.9 (1.5)

Gender

Male 50%

Female 50%

Performance status*

0 43%

1 50%

2 7%

Laboratory

Total bilirubin (mg/dL) 1.7 (1.0)

AST (U/L) 46.9 (4.1)

ALT (U/L) 40.6 (4.4)

CEA level (ng/mL) 411.7 (132.7)

Tumor location

Bilobar 84%

Right lobe 16%

Liver tumor burden

Diameter of index lesion (cm) 7.2 (1.0)

Overall tumor volume (mL) 280.3 (46.0)

Prior chemotherapy 60%

Prior invasive intervention

Radiofrequency ablation 2%

Partial hepatectomy 8%

Hepatic arterial pump 2%

Data are presented as mean (SE) or percentage of study

population.

ALT ¼ alanine aminotransferase; AST ¼ aspartate aminotrans-

ferase; CEA ¼ carcinoembryonic antigen.nEastern Cooperative Oncology Group performance status.

was performed to calculate LSF, reflecting the amountof intratumoral shunting (Fig 1a,b) (8). No patients wereexcluded.Cross-sectional imaging was reviewed for the presence

of liver lesions and other metastases before and afterSIRT. The greatest diameter of the largest liver lesion oncross-sectional axial images was measured. Total hepatictumor volume was assessed using three-dimensionalsoftware (TeraRecon; Foster City, California) thatallowed for multiple regions of interest to be createdfor all tumors, with summation of the total tumorvolume. Regions of interest and measurements wereperformed by a board-certified vascular radiology fellow(A.R.D.) and reviewed by a board-certified and Certif-icate of Added Qualification–certified interventionalradiologist (R.O.).The time between SIRT and disease progression and

survival after SIRT was recorded. The medical chartswere reviewed for cancer staging, baseline laboratoryvalues, and chemotherapy agents administered before,during, and after the evaluation before SIRT. Survivalwas determined by chart review; when there was norecord of death, the last note in the chart was used as anestimate of survival. Additionally, a research databasewas reviewed for the presence of genetic mutationsdetermined by SNaPshot analysis of pathology speci-mens obtained by percutaneous biopsy (9).The relationship of LSF to the presence of lung

metastases was assessed with Kaplan-Meier survivalanalysis by dividing patients into two groups by themedian LSF (7.3): patients with lower (o 7.3%; n ¼ 31)and higher (4 7.3%; n ¼ 31) LSF. Similarly, therelationship of LSF to patient survival after SIRT,progression of liver disease, and time to next new distantmetastasis was also characterized. The relationship oftumor burden, assessed by both the greatest diameter ofthe dominant lesion and the overall hepatic tumorvolume, and the presence of lung metastasis and survivalwere assessed by Cox regression analysis. The impactof receiving chemotherapy before, during, and afterthe evaluation before SIRT was evaluated using Wil-coxon signed rank sum test. Finally, whether tumorburden or LSF differed among patients with and withoutgenetic mutations was assessed using Mann-WhitneyU tests.

RESULTS

The study included 62 patients with liver metastasesfrom colorectal cancer (31 women and 31 men; meanage, 63 y, SD 12) (Table). Documented lung metastasesin addition to liver metastases were present in 33patients. Patients with higher LSF were significantlymore likely to have lung metastases at the time of theevaluation before SIRT (P ¼ .0008). This relationshipwas not true for index lesion size or overall liver tumor

Figure 1. Examples of technetium-99m-labeled macroaggregated albumin scans during evaluation before SIRT. After technetium-

99m-labeled macroaggregated albumin is injected into the hepatic artery to be treated during SIRT, the patient is transferred to nuclear

medicine for single photon emission computed tomography and planar imaging to calculate the LSF. (a) In patients with little

intratumoral vascular shunting (4%), activity predominates in the liver (arrow), with little or no activity in the lung (arrowhead). (b) In

patients with a great deal of shunting (12%), lung activity (arrowhead) is quantified and compared with liver activity (arrow) to calculate

the LSF, which determines the safe dose of radioembolization particles that may be administered during SIRT.

Figure 2. Kaplan-Meier survival curve for survival after SIRT.

After initial overlap, patients with lower LSF survive significantly

longer than patients with higher LSF.

Deipolyi et al ’ JVIR1606 ’ LSF in Colorectal Liver Tumors

volume. Higher LSF was not associated with hepatictumor volume (P ¼ .40) or with increased index tumorlesion size (P ¼ .72).Kaplan-Meier analysis demonstrated that patients

with higher LSF had significantly decreased survivalcompared with patients with lower LSF (P ¼ .03). Afterinitial overlap in the two groups’ survival curves, overtime patients with lower LSF lived significantly longer(Fig 2). Patients with lower hepatic tumor volume didnot have longer survival by Cox regression analysis(P 4 .05). Higher LSF did not predict time to thenext new documented metastasis or to the nextdocumented liver lesion (P 4 .05).There were 19 patients who received chemotherapy

before the evaluation before SIRT; 1 patient receivedchemotherapy before and at the time of the evaluationbefore SIRT; 12 patients received chemotherapy beforeand after the evaluation before SIRT; 3 patients receivedchemotherapy after the evaluation before SIRT; and5 patients received chemotherapy before, during, andafter the evaluation before SIRT. There were 22 patientswho received no chemotherapy. In patients who receivedchemotherapy, multiple drug regimens were used. Of the37 patients who received chemotherapy before SIRT, 20received cetuximab; 28 received irinotecan; 29 receivedbevacizumab; 22 received oxaliplatin; 34 received5-fluorouracil; 21 received leucovorin; 7 received pan-itumumab; 10 received a study drug; and 1 receivedcarboplatin, paclitaxel, and cisplatin. Of the six patientsreceiving chemotherapy concurrent with SIRT, two rec-eived cetuximab, three received irinotecan, two receivedbevacizumab, four received 5-fluorouracil, one receivedcisplatin, and one received panitumumab. Of the 20

patients who received chemotherapy after SIRT, 8received cetuximab, 9 received irinotecan, 10 receivedbevacizumab, 5 received oxaliplatin, 7 received 5-fluo-rouracil, 1 received leucovorin, 5 received panitumumab,5 received a study drug, 1 received gemcitabine, and 2received mitomycin C. Kaplan-Meier survival analysisdemonstrated that any chemotherapy during and afterSIRT had no impact on survival. Patients who receivedany chemotherapy before the evaluation before SIRTtended to have lower LSF (P ¼ .08). Patients whoreceived chemotherapy before the evaluation beforeSIRT and had low LSF had the longest survival afterSIRT (P ¼ .02).SNaPshot analysis of liver biopsy specimens from 16

patients was performed. Nine patients had cancer

Volume 25 ’ Number 10 ’ October ’ 2014 1607

mutations (APC, BRAF, EGFR, IDH1, KRAS, TP53),and no mutations were identified in seven. Patients withmutations had significantly higher LSF by Mann-Whitney U test (mean, 11.0; SD, 3.8) compared withpatients without mutations (mean, 7.3; SD, 0.9) (U ¼ 7;Z ¼ 2.54; P ¼ .01). Patients with genetic mutations didnot have increased hepatic tumor volume or index lesionsize (P 4 .05).

DISCUSSION

Although metastasis is the leading cause of cancer-related deaths, its pathogenesis is poorly understood(2). The findings of this study suggest that intratumoralvascular shunting, quantified as LSF, is associated withdistant metastasis and decreased survival. Although LSFpredicted these outcomes, hepatic tumor volume andindex lesion size did not. It is possible that specifictumoral features including shunting may better predictmetastatic potential than the burden of malignantdisease as indicated by overall tumor volume or indexlesion size. Correlation between LSF and tumor size hasbeen shown previously (10). However, the present studysuggests that LSF may be a novel marker of metastasisand survival.Patients with cancer mutations demonstrated more

intratumoral shunting than patients without mutations.This study implicated various genes, including APC, atumor suppressor protein; BRAF, a protein kinase thatregulates cell division and differentiation; EGFR, a cellsurface receptor that regulates cell proliferation; IDH1,isocitrate dehydrogenase, an enzyme involved in cellularmetabolism; KRAS, an oncogene that promotes tumorcell invasiveness; and TP53, a tumor suppressor proteinthat regulates cell cycle arrest, apoptosis, DNA repair,and metabolism (11).It is unclear how these mutations relate to intra-

tumoral shunting, which could partly be due to aproangiogenic tumor microenvironment (12). Some ofthe pathways highlighted in this study possibly relate toangiogenesis, mediated by multiple mechanisms. Arte-riovenous shunts result in poor delivery of nutrients tocapillaries supplying tumors, inducing further angio-genesis (13). Such shunts could interfere with del-ivery of therapies including chemotherapy, SIRT, anddrug-eluting beads in transarterial chemoembo-lization (14). Understanding the etiology of shunts hasimportant implications for targeted cancer therapy.Further research is necessary to delineate the role thatthese mutations play in the creation of intratumoralshunts; studying other vascular tumors and arterio-venous malformations could provide insight into thepathogenesis of metastasis and possibly therapeutictargets.A significant interaction occurred in which patients

who received chemotherapy before the evaluation before

SIRT and had low LSF had significantly longer survival.There was a trend for chemotherapy administered beforethe evaluation before SIRT to be associated with lowerLSF. It is possible that chemotherapy may reduce theamount of intratumoral vascular shunting. One theory isthat certain chemotherapeutic agents, such as vascularendothelial growth factor inhibitors, may improve out-comes by “normalizing” tumor vascularity (15). Vesselssupplying tumors are heterogeneous and abnormal,characterized by a wide variety of subtypes, includingleaky sinusoids, capillaries, and arteriovenous shunts,leading to hypoperfusion of tumor parenchyma, whichdrives disorderly angiogenesis (15). Although counter-intuitive, normalizing such vasculature and improvingblood flow to tumors may hypothetically decrease theproangiogenic drive and improve perfusion and efficacyof chemotherapies and other treatments. In one study,patients with hepatocellular carcinoma and excessiveintratumoral shunting before the evaluation beforeSIRT, precluding SIRT, received sorafenib, a smallmolecular inhibitor of several tyrosine protein kinases(eg, vascular endothelial growth factor and platelet-derived growth factor receptors). Sorafenib treatmentreduced intratumoral shunting and LSF, allowing foreventual treatment with SIRT (16). Theoretically, adju-vant chemotherapy could normalize tumor vasculature,decreasing vascular shunting and possibly promotingefficacy of SIRT and other liver-targeted therapies.The primary limitation of this study is its retrospective

nature, precluding the assessment of the impact ofincreased LSF on progression of metastatic disease.Future studies could assess the relationship betweenhigh LSF and the development of distant metastases.Another limitation is that patients were treated withmultiple chemotherapeutic agents, precluding the abilityto evaluate the relative contribution of specific chemo-therapies. Because of the retrospective nature of thestudy, it is possible that there is selection bias in patientswho did and did not receive chemotherapy, which couldconfound the relationship of chemotherapy, LSF, andsurvival. Finally, the small number of patients in whomSNaPshot analysis was performed precludes assessmentof the relationship of specific cancer genes to LSF.Future work may focus on the impact of particularpathways.In conclusion, LSF may be a novel marker of meta-

stasis and cancer survival and may be more predictivethan tumor size. Increased LSF secondary to vascularshunting within liver metastases indicated the presenceof distant lesions. Intratumoral shunting could theoret-ically provide a conduit for circulating tumor cellsto access more remote organs, bypassing filtration byliver parenchyma, and could be an important factorin metastatic potential of colorectal cancer. Certaingenetic mutations may promote the formation of thesevascular shunts and could represent possible therapeutictargets.

Deipolyi et al ’ JVIR1608 ’ LSF in Colorectal Liver Tumors

ACKNOWLEDGMENT

This study was supported by National Institutes ofHealth grant RO3 CA172738 (R.O.) and the Depart-ment of Radiology, Massachusetts General Hospital.

REFERENCES

1. Sheth RA, Hesketh R, Deipolyi AR, Oklu R. Circulating tumor cells:personalized medicine in interventional oncology? J Vasc Interv Radiol2012; 24:221–228.

2. Stein U, Schlag PM. Clinical, biological, and molecular aspects ofmetastasis in colorectal cancer. Cancer Res 2007; 176:61–80.

3. Nguyen DX, Massague J. Genetic determinants of cancer metastasis.Nat Rev Genet 2007; 8:341–352.

4. Yu M, Stott S, Toner M, Maheswaran S, Haber DA. Circulating tumorcells: approaches to isolation and characterization. J Cell Biol 2011; 192:373–382.

5. Bookstein JJ, Cho KJ, Davis GB, Dail D. Arterioportal communications:observations and hypotheses concerning transsinusoidal and transvasaltypes. Radiology 1982; 142:581–590.

6. Ngan H, Peh WC. Arteriovenous shunting in hepatocellular carcinoma:its prevalence and clinical significance. Clin Radiol 1997; 52:36–40.

7. Salem R, Thurston KG. Radioembolization with 90Yttrium micro-spheres: a state-of-the-art brachytherapy treatment for primary and

secondary liver malignancies. Part 1: technical and methodologic consid-erations. J Vasc Interv Radiol 2006; 17:1251–1278.

8. Ahmadzadehfar H, Sabet A, Biermann K, et al. The significance of99mTc-MAA SPECT/CT liver perfusion imaging in treatment planning for90Y-microsphere selective internal radiation treatment. J Nucl Med 2010;51:1206–1212.

9. Dias-Santagata D, Akhavanfard S, David SS, et al. Rapid targetedmutational analysis of human tumours: a clinical platform to guidepersonalized cancer medicine. EMBO Mol Med 2010; 2:146–158.

10. Ho S, Lau WY, Leung WT, et al. Arteriovenous shunts in patients withhepatic tumors. J Nucl Med 1997; 38:1201–1205.

11. Gene. Available at: http://www.ncbi.nlm.nih.gov/gene. Accessed June15, 2014.

12. Weis SM, Cheresh DA. Tumor angiogenesis: molecular pathways andtherapeutic targets. Nat Med 2011; 17:1359–1370.

13. Pries AR, Hopfner M, le Noble F, Dewhirst MW, Secomb TW. Theshunt problem: control of functional shunting in normal and tumourvasculature. Nat Rev Cancer 2010; 10:587–593.

14. Sheth RA, Hesketh R, Kong DS, Wicky S, Oklu R. Barriers to drugdelivery in interventional oncology. J Vasc Interv Radiol 2013; 24:1201–1207.

15. Carmeliet P, Jain RK. Principles and mechanisms of vessel normal-ization for cancer and other angiogenic diseases. Nat Rev Drug Discov2011; 10:417–427.

16. Theysohn JM, Schlaak JF, Muller S, et al. Selective internal radiationtherapy of hepatocellular carcinoma: potential hepatopulmonary shuntreduction after sorafenib administration. J Vasc Interv Radiol 2012; 23:949–952.