Dr. Gary J. Becker Young Investigator Award: Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer: Activity, Radiographic Response, and Survival

Special Communication

Dr. Gary J. Becker Young Investigator Award: Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer: Activity, Radiographic Response, and Survival Daniel Y. Sze, MD, PhD, Scott M. Freeman, MD, Suzanne M. Slonim, MD, Shaun L. Samuels, MD, James C. Andrews, MD, Marshall Hicks, MD, Kamran Ahrar, MD, Sanjay Gupta, MD, and Tony R. Reid, MD, PhD

PURPOSE: To determine the antitumoral activity and radiographic response pattern of intraarterial administration of a selective replication-competent adenovirus in patients with hepatic metastases from gastrointestinal carcinomas. MATERIALS AND METHODS: Thirty-five patients were treated, seven in the dose-escalation phase and 28 at high doses. Inclusion criteria allowed mild laboratory value and performance status abnormalities and as much as 50% replacement of hepatic volume by tumor. An attenuated adenovirus that selectively replicates in p53-deficient cells (Onyx-015) was administered by hepatic arterial infusion at doses as high as 2 ⴛ 1012 particles for two cycles. Subsequent cycles (maximum of eight total) were administered in combination with intravenous 5-fluorouracil (5-FU) and leucovorin. RESULTS: Tumor responses were demonstrated in combination with chemotherapy, even in 5-FU–resistant patients. The 15 patients who responded radiographically showed a pattern of acute tumor enlargement despite normalization of laboratory and clinical parameters, followed by very slow regression of tumor size. Radiographic response did not correlate with p53 status. Median survival of radiographic responders (475 days) was significantly longer than that of nonresponders (143 days). CONCLUSIONS: Hepatic arterial infusion of the replication-selective adenovirus Onyx-015 in combination with chemotherapy resulted in tumor regressions in select patients, including some in whom previous chemotherapy had failed. A biphasic radiographic response pattern was demonstrated. The mechanism of action appears to be more complex than that seen in vitro. Index terms:

Chemotherapeutic infusion

•

Gene therapy

•

Liver neoplasms, therapy

•

SCVIR Annual Meeting, 2002

•

Viruses

J Vasc Interv Radiol 2003; 14:279 –290 Abbreviation:

CEA ⫽ carcinoembryonic antigen, 5-FU ⫽ 5-fluorouracil

EMERGING therapies for cancer are providing new hope for patients and their physicians. New pharmaceutical

agents have resulted in small but real improvements in outcomes, and great enthusiasm has developed for gene

Fom the Division of Cardiovascular and Interventional Radiology (D.Y.S., S.M.S., S.L.S.), Stanford University Medical Center, Stanford, and Palo Alto Veterans Administration Medical Center, Palo Alto, California; Onyx Pharmaceuticals, Inc. (S.M.F.), Richmond, California; Department of Radiology (J.C.A.), Mayo Clinic, Rochester, Minnesota; Department of Radiology (M.H., K.A., S.G.), University of Texas M.D. Anderson Cancer Center, Houston, Texas; and Division of Medical Oncology (T.R.R.), Palo Alto Veterans Administration Medical Center and Stanford University Medical Center, Palo Alto, California. From the SCVIR 2002 Annual Meeting. Received June 10, 2002; revision requested August 12; revision received November 4; accepted Novem-

ber 12. Address correspondence to D.Y.S., Cardiovascular and Interventional Radiology, Stanford University Medical Center, H-3646, Stanford, CA 94305-5642; E-mail: [email protected] D.Y.S. and T.R.R. are current paid consultants to Onyx Pharmaceuticals, Inc.; S.M.F. is an employee of Onyx Pharmaceuticals, Inc. None of the other authors has identified a potential conflict of interest. © SIR, 2003 DOI: 10.1097/01.RVI.0000058422.01661.1E

therapy and molecular medicine as possible sources of agents with potentially novel and revolutionary mechanisms (1–3). The majority of gene-therapy research has focused on the use of live but replication-incompetent viruses as vehicles to carry therapeutic genes into tumor cells (4). However, insufficient levels and durability of gene expression have limited this approach. In addition, the death in 1999 of a teenage patient undergoing intraarterial adenoviral gene therapy for ornithine transcarbamylase deficiency alerted the medical community and the public to safety issues concerning the use of live viruses as therapy, lead-

279

280

•

Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer

ing to suspension of nearly all genetherapy trials in the United States and Europe (5). A parallel strategy with use of replication-competent but attenuated viruses has emerged as a promising alternative (6 – 8). Rather than acting as carriers of genetic payloads, these viruses are engineered to be directly and selectively cytotoxic to tumor cells. The first-generation agent dl1520, also known as Onyx-015 (Onyx Pharmaceuticals, Richmond, CA), is a replication-selective adenovirus type 2/5 chimera with the E1b-55kD gene deleted (9). E1b is an adenoviral product that binds to and inactivates the human p53 tumor suppressor, enabling the virus to infect a human host successfully and cause the common cold. In theory, with the E1b gene deleted, Onyx-015 is ineffective at replicating in and killing normal cells but is fully infectious in p53deficient cells, which include the majority of human solid malignancies (10). Preclinical studies have confirmed the retained infectivity of Onyx-015 in p53 mutant cells and decreased pathogenicity in normal cells when compared to wild-type adenovirus (11). In light of these data and drawing on the extensive experience of intraarterial chemotherapy in the treatment of hepatic malignancies, we performed a phase-I/II dose-escalation human clinical trial of administration of Onyx-015 into the hepatic artery to treat gastrointestinal (primarily colorectal) carcinomas metastatic to the liver. Earlier publications have addressed feasibility, safety, pharmacokinetics, and immune response of this trial (12). The current report focuses on the radiographic response and substratifies the data according to this response.

MATERIALS AND METHODS This trial was overseen by the U.S. Food and Drug Administration. Patient protocol was approved by the institutional review board of each participating hospital. Clinical toxicity and efficacy data were monitored by an independent clinical trial consulting organization.

Inclusion Criteria • Histologically or cytologically confirmed adenocarcinoma of gastrointestinal origin; • Surgically unresectable disease; • Patent hepatic artery supplying entire liver and all intrahepatic tumor masses; • Karnofsky performance status ⱖ70%; • Life expectancy ⱖ3 months; • Age ⱖ18 years; • Infertility or current use of reliable contraception if female; • Creatinine level ⬍2.0 mg/dL, aspartate aminotransferase and alanine aminotransferase levels less than three times the upper limit of normal, total bilirubin level ⬍2.0 mg/dL, International Normalized Ratio ⬍2.0, partial thromboplastin time within normal limits, neutrophil count ⬎1,500/mL, hemoglobin level ⬎9 g/dL, and platelet count ⬎100,000/mL. Exclusion Criteria • Cirrhosis, chronic active hepatitis, or other chronic liver dysfunction; • ⬎50% replacement of liver volume by tumor, estimated from computed tomography (CT); • History of variceal bleeding within preceding 8 weeks (as a sign of portal hypertension or obstruction); • Active infection, including with HIV; • Viral syndrome within preceding 2 weeks; • Chemotherapy within preceding 3 weeks (6 weeks for nitrosourea or mitomycin-C); • Radiation therapy to target tumor(s) within preceding 4 weeks; • Concomitant hematologic malignancy; • Chronic immunosuppression; • Pregnancy or lactation; • Previous administration of adenoviral vectors; • Participation in any other investigational protocol within preceding 4 weeks.

March 2003

JVIR

Patient Enrollment Earlier therapy for primary or metastatic tumor was not defined in the eligibility criteria, but the large majority of patients had undergone earlier failed systemic chemotherapy regimens. Seven patients were enrolled in the dose-escalation portion of phase I. Twenty-eight patients were enrolled in the high-dose portion of phase I and in phase II. Patients underwent image-guided biopsy of a target lesion and of normal-appearing liver in addition to contrast-enhanced CT in the 2 weeks before commencement of viral therapy. p53 gene status of the tumor did not affect enrollment eligibility, but gene status was assessed by sequencing of exons 2–11 and/or immunostaining when sufficient amounts of tissue could be obtained from the biopsy. Biopsy of a target lesion and of surrounding normal liver was repeated 72 hours after initial viral infusion to evaluate for necrosis and evidence of viral replication as well as for cytopathic effects on surrounding hepatocytes. Biopsies were either ultrasoundor CT-guided, and efforts were made to obtain tissue within 1 cm of the margin to avoid central necrosis. Viral Infusions Viral infusions were performed on days 1 and 8 (cycles 1 and 2) to determine the safety of Onyx-015 as a single agent (Table 1). Infusion was repeated on days 22 and 50 (cycles 3 and 4), each followed by intravenous chemotherapy within 6 hours of viral infusion. After these first four infusions, as many as four additional optional infusions could be administered at 28-day intervals, depending on individual patient tolerance and response. Intravenous chemotherapy was administered after viral infusions in cycles 3 and thereafter to test for possible synergy or chemosensitization. Intravenous 5-fluorouracil (5-FU) was given as a 425-mg/m2/d bolus dose for 5 consecutive days, along with leucovorin at a dose of 20 mg/m2/d (Mayo regimen [13]). Dose adjustments of chemotherapeutic agents were made as needed, with successive administrations according to patient tolerance and toxicity, but viral dose was not adjustable. Complete hepatic arteriography

Volume 14

Number 3

Sze et al

•

281

Table 1 Treatment and Imaging Scheme Procedures Treatment Viral infusion 5-FU/leucovorin† Diagnostic Blood tests‡ CT Biopsy§

Pre*

Day 1

Day 3

Day 8

Day 15

Day 22

Day 29

Day 50

Day 58

– –

X –

– –

X –

– –

X X

– –

X X

– –

X X X

– – –

– – X

X – –

X – –

X – –

X X –

X – –

X X –

* Pre-enrollment studies were performed within 14 days of the first viral infusion. † 5-FU/leucovorin treatments entailed 5 days of intravenous administration (Mayo protocol). ‡ Blood tests included complete blood count, chemistry values, CEA, lactate dehydrogenase, liver function tests, and coagulation parameters. § Biopsies included sampling of target lesion and of normal liver parenchyma. Note.—At day 78 and every 28 days thereafter, complete blood tests, CT, repeat hepatic arteriography, and cycles 5– 8 viral infusions at discretion of investigation team were performed.

was performed on each patient before viral infusion to define arterial anatomy and to confirm arterial supply to intrahepatic tumor(s). In patients with accessory or replaced hepatic arteries, the proportion of the liver supplied by each artery was estimated from the parenchymal phase of the arteriogram and the viral dose was divided proportionally. After optimal positioning of the catheter to ensure minimal perfusion of and reflux into arteries not supplying the liver, a single infusion of 10 mL viral solution was performed over a period of 3 minutes, followed by flushing with 10-mL physiologic electrolyte solution (5% dextrose in water/electrolyte 48) given at the same rate and site. One patient already had a surgically implanted arterial pump for continuous arterial infusion chemotherapy; function and complete hepatic perfusion had been confirmed by injection of technetium sulfur colloid. This pump was accessed percutaneously per manufacturer instructions for this patient’s viral infusions. All procedures were performed under conscious sedation with fentanyl and midazolam and cardiac and blood pressure monitoring. Patients were given oral acetaminophen immediately after infusion and as needed thereafter. Narcotic analgesics and antiemetics were provided as needed. The viral dose was specified for each cohort. Frozen vials of viral solution were warmed and diluted with normal saline solution to the specified titer for each patient cohort. Dilutions were performed in Biologic Safety

Level 2 cabinets. Viral solutions were handled at a temperature of approximately 5°C during dilution and transport and were warmed to room temperature (20°C) immediately before infusion. Doses started at 2 ⫻ 108 particles, corresponding to 1 ⫻ 107 plaque-forming units per infusion, and increased by a half log-unit with each cohort to a maximum of 2 ⫻ 1012 particles (1 ⫻ 1011 plaque-forming units) per infusion. Therefore, nine cohorts were specified, ranging from 107 to 1011 plaque-forming units per infusion. Clinical Assessment Blood was drawn before each treatment, on days 2 and 4 after the first infusion, and on day 8 after each infusion. Serum chemistry, blood count, coagulation parameters, liver function tests, and lactate dehydrogenase levels were assayed. In addition, serum carcinoembryonic antigen (CEA) was assayed at the time of the first and third infusions and monthly thereafter. Signs of clinical toxicity were assessed throughout the treatment protocol period and for 28 days after completion of the protocol as described in detail elsewhere (12). Pharmacokinetic and immunologic parameters were measured as described in detail elsewhere (14). Briefly, viral genomes were assayed by polymerase chain reaction of circulating blood to determine elimination kinetics and later to document evidence of viral replication. Circulating

antibodies were also assayed before and after viral therapy by incubating plasma with adenoviral stock to detect neutralization. Circulating proinflammatory cytokines, including interleukins 1, 6, and 10, tumor necrosis factor, and interferon gamma, were assayed by polymerase chain reaction of gene expression in peripheral mononuclear cells and by enzyme-linked immunosorbent assay of serum. Radiographic Assessment A baseline contrast-enhanced CT scan was obtained within the 14 days preceding the initial viral infusion. CT was repeated on a monthly basis immediately before infusions for the duration of the treatment period. When possible, additional CT scans were obtained after completion of the protocol at the discretion of the referring oncologist. Portal venous-phase images were obtained as per the standard protocol of each participating institution. The segmental location and size of each discrete lesion was recorded by measurement of the longest axis and perpendicular axis of each measurable lesion. Lesions outside the liver were also recorded but not included in the total target hepatic tumor burden. Total burden was estimated according to oncologic literature standards as a summation of cross-sectional areas of individual lesions. For illustration purposes only, the estimated tumor volumes were also calculated, assuming ellipsoid shapes, with the geometric mean of the first two measure-

282

•

Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer

Table 2 Baseline Patient Characteristics by Dose Group Characteristic Mean age (y) ⫾ SD Sex Male Female Baseline Karnofsky performance status 100 90 80 70 Ethnicity White Black Asian Other Earlier therapy None Surgery only Chemotherapy only Surgery/Chemotherapy Surgery/Chemotherapy/other Histology Colorectal adenocarcinoma Other Gastric Pancreatic Total tumor burden (sum of cross-sectional areas, cm2) Mean Median Range Tumor p53 status Wild-type Mutant Not evaluable

Dose-escalation (n ⫽ 7)

High-dose (n ⫽ 28)

60.7 ⫾ 6.1

60.4 ⫾ 10.7

6 (86) 1 (14)

20 (71) 8 (29)

4 (57) 2 (29) 1 (14) 0

15 (54) 3 (11) 8 (29) 2 (7)

6 (86) 1 (14) 0 0

25 (89) 0 3 (11) 0

0 0 1 (14) 5 (71) 1 (14)

0 3 (11) 1 (4) 21 (75) 3 (11)

5 (71) 2 (29) 1 (14) 1 (14)

26 (93) 2 (7) 1 (4) 1 (4)

131.6 119.0 46–205

144.6 91.5 7–482

4 1 2

12 11 5

Note.—Values in parentheses are percentages.

ments used as the third dimension. Other pertinent findings, including ascites, pleural effusion, biliary ductal dilation, and body habitus were also recorded. All quantitative measurements were performed by one radiologist for consistency, and patients were identified by subject numbers only. A random sample of 10 scans were reread to calculate intraobserver variability and total tumor burdens were found to vary by an SD of 5%. World Health Organization standards of radiographic tumor response were used. Complete response was defined as complete disappearance of all target lesions. Partial response was defined as a decrease in total crosssectional tumor burden of 50%–99%. Minor response was defined as a decrease of 25%– 49%. Increase of total intrahepatic tumor burden of less than

25% was classified as stable disease and increase of more than 25% was classified as progressive disease. Because low attenuation is traditionally associated with necrosis, the attenuation and distribution of pixel values of the largest target lesions were measured when feasible. There is no accepted standard method to quantify the degree of necrosis of tumors, so the dimensions of the entire masses were used without correction for heterogeneity and suspected necrosis. A tailored definition of “radiographic response” was also created exclusively for the specific characteristics of this trial. Because of the unprecedented pattern of tumor response, because the degree of necrosis was impossible to assess, and because regression of tumor size was incongruously slow, we also performed

March 2003

JVIR

subset analyses to stratify patients into a group in which at least a 10% (2 ⫻ SD of intraobserver variability) decrease from the maximum total hepatic tumor burden was exhibited (“responders”) and a group in which even such a modest decrease was not evident (“nonresponders”). Statistical Analyses Kaplan-Meier survival curves were calculated for radiographic responders and nonresponders and compared with use of a log-rank test (␹2 with one degree of freedom). Correlation of p53 genotype with radiographic response was analyzed with use of a Pearson ␹2 test with one degree of freedom. The SPSS 11.0.1 statistical package (SPSS, Chicago, IL) was used for all calculations.

RESULTS Patients Patient demographics are listed in Table 2. The large majority of patients had undergone previous therapies, including surgery and intravenous chemotherapy. Only 15% of enrolled patients were chemotherapy-naive. The mean and median total cross-sectional tumor burdens were 142 cm2 and 112 cm2 (range, 7– 482 cm2), respectively, corresponding to estimated mean and median tumor volumes of 508 mL and 281 mL (range, 7–1,917 mL). Preinfusion biopsies yielded adequate material for p53 sequencing or immunostaining in 28 patients, which revealed mutations in 12 (43%). Biopsies Nearly all postviral infusion biopsy samples of target lesions yielded predominantly necrotic debris with negligible viable cells. Of course, sampling artifact cannot be excluded, but biopsies were obtained from the peripheral 1-cm rim of lesions. Three patients underwent extensive histopathologic evaluation of hepatic parenchyma after viral infusion therapy (one after a hepatic lobectomy, two at autopsy). In these three patients, hepatic tissue was normal without evidence of fibrosis, steatosis, cholestasis, vasculitis, or leukocytic infiltration. In the autopsy sub-

Volume 14

Number 3

Sze et al

•

283

jects, all extrahepatic tissues including lung and bone marrow also appeared normal. In the patient who underwent hepatic lobectomy, the residual 2.6-cm ⫻ 2.0-cm lesion was almost completely filled with a mucinous substance with rare small clusters of viable tumor cells. Viral Infusions All infusions were completed successfully. The use of microcatheters was necessary in approximately 25% of patients because of variant hepatic arterial anatomy, celiac or hepatic arterial obstructions, or excessive reflux into the right gastric or gastroduodenal artery. Seventeen percent of patients had replaced right hepatic arteries, 7% had replaced left hepatic arteries, 3% had replaced proper hepatic arteries, 3% had aberrant left hepatic arteries originating from the splenic artery, and 9% had celiac artery occlusions. Follow-up arteriograms in later viral infusion cycles revealed no evidence of vasculitis, aneurysms, or other vascular abnormalities. Despite undergoing as many as eight femoral arterial punctures in 7 months, no patient had major accessrelated complications or required transfusion or surgery. Infusions via the surgically implanted pump in one patient were also successful, with no evidence of procedural complications. An asymptomatic small partial splenic infarct, probably an embolic complication of an angiographic procedure, developed in one patient, who was withdrawn from the study after two infusions in accordance with instructions from the monitoring consultants. Despite the lack of evaluable material in the postinfusion biopsies, other evidence was gathered that suggested that intraarterial administration accomplished successful infection. Viral genomes disappeared from peripheral blood within 6 hours of infusion, but reappeared 72 hours later (approximately one viral generation) in 60% of the patients who received the highest dose (14). Clinical symptoms of infection, including fever, chills, myalgia, and rigors, were seen in more than 90% of patients, and symptoms lasted as long as 1 week. These symptoms, changes in laboratory test values, delayed reappearance of viral genomes, and measured pharmacokinetics did

Figure 1. Changes in angiographic appearance of tumors during the protocol. (a) Selective arteriogram of a replaced right hepatic artery in a 76-year-old patient shows attenuation and stretching of intrahepatic vessels by enlarging tumors. (b) After a 50% decrease in total cross-sectional tumor burden, vessels have become redundant and tortuous. (c) Parenchymal phase of selective left hepatic arteriogram in a 39-year-old patient with a replaced right hepatic artery shows marked hypervascularity of lesions. Note that the gastroduodenal artery has been displaced to the left of midline by the massively enlarged liver. (d) After seven cycles of viral infusions and chemotherapy, hypervascularity is decreased and some tumors appear as voids.

not differ significantly between the first and subsequent infusions. This is despite preexisting antibodies to adenovirus found in as much as 90% of the general population (15) and large increases in titers of neutralizing antibodies measured in our patients after infusions (14), suggesting that even high levels of antibodies can be

avoided or saturated by intraarterial delivery. Of the 35 patients, only 23 completed the four-infusion protocol. The other 12 were withdrawn after one (n ⫽ 1), two (n ⫽ 6), or three (n ⫽ 5) infusions because of lack of tumor shrinkage. Thirteen patients received exactly four infusions. Ten patients re-

284

•

Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer

March 2003

JVIR

Antitumor Activity

Figure 2. Sequential CT scans of a 67-year-old woman who had received previous intravenous 5-FU/leucovorin and irinotecan and continued to progress. CT images at the level of the largest hypoattenuating lesion in segment 7 before commencement of viral treatment (a), 4 weeks later (after two viral infusions alone and one viral infusion in combination with intravenous 5-FU/leucovorin) (b), at 8 weeks (c), and at 14 weeks (d) show initial enlargement of this mass, with increased heterogeneity, increased margination, and decreased attenuation, followed by progressive diminution and homogeneous hypoattenuation. The porta hepatis nodal mass, not seen on all of these images, did not change substantially in size during this period. This patient survived 372 days.

ceived additional infusions beyond the original four, which were prescribed because of encouraging results after the first four infusions. Mean and median numbers of infusions were 4.2 and 4.0, respectively. As is typical of metastatic gastrointestinal carcinomas, the large majority of identified lesions were angiographically hypovascular but had hypervascular rims. Over the course of the protocol, subtle changes in angiographic appearance were noted in several patients, including decreasing hypervascularity of lesion rims and, in patients whose lesions regressed, relief of attenuation and splaying of arterial vasculature (Fig 1).

Toxicity As previously reported (12), intraarterial infusion of Onyx-015 was well tolerated as a single agent (first two dose cycles for each patient) and in combination with intravenous 5-FU and leucovorin (subsequent dose cycles). The maximum tolerated dose was not reached at the highest feasible dose of 1 ⫻ 1011 plaque-forming units per infusion, but nearly all patients reported flulike symptoms. Transient elevations of serum bilirubin and/or transaminase levels observed in approximately one third of patients, suggesting mild hepatotoxicity, resolved spontaneously in 7–14 days.

This study was not specifically designed to assess antitumor activity of Onyx-015, but was primarily a safety trial. Antineoplastic efficacy is difficult to prove in such a nonrandomized study because patients were enrolled with a wide range of previous therapies, tumor burdens, and clinical prognoses. However, evidence was gathered that suggested antitumor activity, even in patients in whom earlier 5-FU/leucovorin chemotherapy regimens had failed. Of the 28 patients in the high-dose cohorts, five minor responses were seen in the 3 months of the defined protocol (18%). Nine patients progressed (32%) and the remaining 14 had stable disease (50%). Over longer-term follow-up, three patients eventually showed partial responses (11%). However, one of these patients was chemotherapy-naive, so the contribution of Onyx-015 to this response was not evaluable. In the dose-escalation cohorts, six patients (86%) exhibited progressive disease and only one patient exhibited stable disease. After longer follow-up, one patient with progressive disease did show enough regression to be reclassified as having stable disease (28%). The paucity of major responses (no complete responses and only three partial responses) was not congruent with the clinical improvement and decreased serum CEA levels observed in many of these patients. Further analysis revealed that the radiographic pattern of response differed from that expected of standard chemotherapy (Figs 2–5) (16,17). Thirty-three of the 35 patients (94%) exhibited enlarged tumors at 1-month follow-up CT, which was initially interpreted as tumor progression. In fact, the basic four viral infusions were not completed in 12 patients because initial tumor enlargement was interpreted as disease progression. This acute enlargement of tumors within the first month after the first viral infusion was accompanied by subjectively increased margination of lesions, including of smaller lesions previously undetected. Twenty patients (57%) continued to show tumor enlargement in subsequent months, confirming progressive disease. However, tumor burden stabilized and eventually decreased slowly in

Volume 14

Number 3

Sze et al

Figure 3. Sequential CT scans of a 59-year-old man who had previously received bolus and continuous intravenous infusion of 5-FU/leucovorin and continued to progress. CT images at the level of the largest lesion in the right lobe before commencement of viral treatment (a), 3 weeks later (after two viral infusions alone) (b), at 7 weeks (c), and at 11 weeks (d). Note acute response to viral infusion. This patient survived 608 days.

•

285

after reaching a maximum but never returned to initial levels. Of the 18 patients who received high-dose Onyx-015 and completed at least four infusions, 10 (56%) exhibited a radiographic response by this definition. The 15 total patients who exhibited a radiographic response survived a median of 475 days (95% CI: 284 – 666 d) and a mean of 534 days (95% CI: 396 – 672 d), with two censored at 1,084 and 1,088 days. In comparison, the 20 nonresponders survived a median of 143 days (95% CI: 134 –152 d) and a mean of 188 days (95% CI: 128 –249 d) (P ⬍ .0001). Although limited by the design of this study, these figures compare favorably to those of other chemotherapeutic regimens and those of supportive care. The response statistics evaluating survival exceeded expectations, particularly in consideration of the paucity of objective responses as measured by standard criteria. The radiographic response did not correlate with p53 status. In fact, a higher proportion of patients with wild-type genotype (50%) exhibited radiographic responses than did those with p53 mutations (33%), but this did not reach significance in this small series (P ⫽ .38) (Table 3).

DISCUSSION the other 15 patients (43%) (Fig 6). With subsequent shrinking of tumors, the initial enlargement and margination was hypothesized to be inflammation, devascularization, and necrosis, but this will need to be confirmed in future studies. On contrast-enhanced CT scans, the attenuation (in HU; mean ⫾ SD) of the largest target lesion in each patient decreased a negligible amount after 1 month (⫺1.7 HU ⫾ 9.5) and 2 months (⫺1.1 HU ⫾ 10.9). This is despite qualitative increases in visible necrosis and heterogeneity and subsequent low-attenuation homogeneity. Percentage of necrosis was therefore confirmed to be incalculable. Radiographic response lagged behind tumor marker and liver function test responses, with very slow resorption of the presumably necrotic tissue (Fig 7). For instance, in one patient whose CEA level had decreased to undetectable levels, total cross-sectional area of lesions was still 193 cm2, corresponding to an estimated half-

liter of tumor, although each lesion appeared homogeneously low in attenuation and necrotic. A 19-fluorodeoxyglucose positron emission tomographic scan was performed at this time and did not show hypermetabolic foci, but a baseline scan was not available for comparison. Survival curves as a function of radiographic response are shown in Figure 8. For the purpose of this analysis, radiographic responders were defined as the patients who exhibited at least a 10% decrease in total tumor cross-sectional burden from the maximum. By this definition, 13 of the 28 patients (46%) who received high-dose virus were radiographic responders and two of the seven patients (28%) in the dose-escalation study were responders. This included the two patients who had a monotonically decreasing tumor burden, the five whose tumor burden initially increased but then decreased to below initial levels, and the eight whose tumor burden decreased

The exciting possibilities of human gene therapy have been discussed for decades, but remain largely unfulfilled. In 1999, a severe setback occurred when a young patient underwent gene therapy for ornithine transcarbamylase deficiency with a replication-incompetent adenovirus and died of direct complications. Since then, greater scrutiny has been paid to the safety of biologic agents. This study tested the feasibility and safety of intraarterial administration of an attenuated but replication-competent virus. The highest dose given in this study represents more than an order of magnitude less virus than was given to the patient with ornithine transcarbamylase, but the replicationcompetence of Onyx-015 could theoretically result in self-amplification and a much larger and longer-lasting dose (6). In fact, the administered dose of Onyx-015 had a serum half-life of only 10 –14 minutes and was undetectable in peripheral blood within 6 hours, but was again detectable ap-

286

•

Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer

Figure 4. Sequential CT scans of a 76-year-old man (same patient as in Fig 1a,1b) who had not received previous chemotherapy and had multiple rapidly enlarging masses, now estimated to occupy 1.6 L of volume. CT images at the level of the left portal vein before commencement of viral treatment (a), 4 weeks later (after three viral doses including one in combination with 5-FU/leucovorin) (b), at 8 weeks (c), and at 22 weeks (d). Note the apparent increase in number and size of lesions after initial treatments. At 22 weeks, the left portal vein had returned to the right of midline. This patient survived 373 days.

Table 3 Radiographic Response as a Function of p53 Status

Dose-escalation cohorts (ⱕ1 ⫻ 1010 pfu) Wild-type p53 Mutant p53 Status not evaluable Subtotal High-dose cohorts (ⱖ3 ⫻ 1010 pfu) Wild-type p53 Mutant p53 Status not evaluable Subtotal Totals Wild-type p53 Mutant p53 (P ⫽ .38) Note.—pfu ⫽ plaque-forming units.

Radiographic Responders

Nonresponders

Totals

1 1 0 2

3 0 2 5

4 1 2 7

7 3 3 13 13 8 (50%) 4 (33%)

5 8 2 15 20 8 8

12 11 5 28 35

March 2003

JVIR

proximately 72 hours later at concentrations as high as 4 ⫻ 106 genomes per mL blood, signifying successful infection and active replication (14). The safety, efficacy, and pharmacokinetics of new agents that target cells based on genetic characteristics need to be determined to evaluate their utility in treatment of human malignancies. Existing human trial data are limited to protocols involving direct intraparenchymal injection of viral suspensions, either with image guidance for pancreatic adenocarcinomas (18) or by direct palpation and visualization of head and neck tumors (19,20). Although clinical responses to Onyx-015 alone were rare, a possible synergy with standard chemotherapeutic agents was identified in the head and neck. Responses were confined to the regions of direct injection, and applications of this method of delivery in multifocal, microscopic, and inaccessible tumors are limited. As is true of other anticancer agents, the method of delivery of engineered viruses needs to be optimized, and here the techniques of interventional radiology may play an important role. Preclinical studies of other viral vectors have shown that the rate of transfection is a function of the route of administration and that intraarterial administration can have advantages over intravenous, intraportal, and interstitial routes (21–26). In addition, maximal perfusion of tumors versus hepatic parenchyma may be achieved by intraarterial infusion (27). This has been shown even in colorectal carcinoma metastases, which are not typically hypervascular radiographically. Hepatic metastases are responsible for approximately 80% of the 55,000 deaths that occur each year in the United States from colorectal carcinoma, and current chemotherapeutic regimens show response rates of only approximately 20% with little survival benefit (28). However, intravascular administration of live virus is potentially risky, and the dose-limiting toxicity of Onyx-015 injected into the tail veins of a nude mouse model was hepatic necrosis (29). The ability of Onyx-015 to affect tumors remains incompletely understood. For the virus particles to reach tumor cells physically, they must traverse the blood/tumor barrier and the extracellular space, overcome in-

Volume 14

Number 3

Sze et al

Figure 5. Sequential CT scans of a 39-year-old man (same patient as in Fig 1c,1d) who had progressed despite bolus and continuous infusion of 5-FU/leucovorin, presenting with approximately 1.3 L of tumor, abdominal pain, and early satiety. CT images at the level of the left portal vein before commencement of viral treatment (a), 3 weeks later (after two viral infusions alone) (b), at 21 weeks (c), and at 56 weeks (c). After completion of the Onyx-015 protocol, the patient received two doses of irinotecan at approximately 30 weeks, after which the regimen was discontinued because of toxicity. However, contribution of irinotecan to tumor response cannot be excluded. Note resolution of cachexia. This patient survived 783 days.

Figure 6. Tumor burden of the 15 radiographic responders as defined in the text. Total intrahepatic tumor burden is normalized to burden at enrollment in protocol. Thirteen patients exhibited initial tumor enlargement, followed by at least 10% decrease. Only two patients showed initial decreases, of 7% and 1%, respectively. Complete data are available for only the 3 months of the defined protocol.

•

287

creased intratumoral oncotic pressure (30), and bind to the human coxsackieadenovirus receptor, which may be expressed in various amounts on tumor cell surfaces (31). Even macromolecules orders of magnitude smaller than viruses are notoriously difficult to deliver to tumors. Neutralization by antibodies, binding to the wrong cells, inability to escape the vascular milieu, lack of receptors or presence of mutant receptors on the target cells, or insufficient dose could all thwart potential oncolytic activity. Some of these pitfalls can be minimized by the use of locoregional intraarterial administration. The theoretic selectivity of Onyx015 relies on its ability to replicate unchecked in the absence of p53, a mechanism confirmed in vitro and in preclinical studies. Clearly, this is not the sole mechanism of action (32,33). At least eight of the 15 radiographic responders in our study had wild-type p53 genotype, and patients with wildtype genotype survived slightly longer than patients with mutant genotype (14). p53 function is very complex, and even with wild-type p53 genotype, other mutations such as in p14(ARF) can also render the p53 tumor suppressor system ineffective and can render these cells susceptible to Onyx-015 infection (34,35). The acute swelling of tumors reported here, the induction of cytokines, and the disappearance of peripheral lymphocytes from the circulating blood reported elsewhere (14) all suggest at least another mechanism as an immunoadjuvant or chemosensitizer. Whether this synergistic mechanism would be induced by wild-type virus is unknown, but Onyx-015 theoretically confers a degree of safety by not harming normal cells. Use of the adenoviral platform, a common community-acquired infection, has advantages and disadvantages. Viral solutions may be handled at the relatively low Biologic Safety Level 2. In addition, the Onyx-015 virus is attenuated, so it is classified as similar to a vaccine. Reversal of the mutation would result in a wild-type virus that merely causes the common cold. Infection epidemics would be nearly impossible because of the attenuation and the preexisting immunity in the general population. However, the presence of neutralizing antibodies in a patient can prevent transfection

288

•

Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer

March 2003

JVIR

Figure 7. Radiographic response versus CEA response. (a) CEA trends in patients who did not exhibit radiographic response as defined in the text. Arrows denote viral infusions. Of the 20 patients who did not show morphologic tumor shrinkage, only two experienced net decreases in CEA over the protocol period. (b) CEA trends in patients who exhibited radiographic response. Of the 15 patients in whom tumor burden decreased by at least 10%, 10 showed a net decrease in CEA over the protocol period. (c) Progression and regression of serum CEA (squares) and of total cross-sectional area of hepatic lesions (diamonds) are plotted for the 76-year-old patient in Fig 1a,1b and Fig 4. Positron emission tomographic scan performed after the seventh infusion did not reveal hypermetabolic foci despite a large amount of residual masses.

and therapeutic efficacy when virus is administered intravenously or otherwise systemically. Administration of Onyx-015 into the hepatic artery apparently saturated and overwhelmed the available antibodies in the limited hepatic vascular space. The advantage of overcoming existing immunity with the use of an arterial route of administration has also been demonstrated in animal models with herpes virus (26). Although the intraarterial route of administration does appear to have advantages, monthly arterial punctures could become problematic, particularly in an elderly patient population with peripheral vascular disease.

Systemic intravenous administration, although potentially more economical, would require doses that are orders of magnitude greater to overcome specific and nonspecific binding and antibody neutralization. Implanted pumps or ports may provide a safe alternative in the future in selected patients. The statistics for objective tumor response according to traditional definitions were somewhat disappointing, with zero complete responses and five minimal responses (18% of the highdose cohort) within the first 3 months of the protocol and only three partial responses (11% of the high-dose co-

hort) with longer follow-up. However, the pattern of lesion evolution and resorption was atypical, involving acute enlargement and heterogeneity, followed by more uniform hypoattenuation and very slowly decreasing size. For the large majority of the radiographic responders, tumor shrinkage did not occur until after three viral infusions, and even then the cumulative net change was negative for only half of these patients. Twelve patients were suspended from the protocol before completing the basic four infusions because of lack of tumor shrinkage, and it is interesting to consider whether these patients would have

Volume 14

Number 3

Sze et al

Figure 8. Survival curves for patients treated with Onyx015, comparing those who exhibited a radiographic response as defined in the text (n ⫽ 15) to those who did not (n ⫽ 20). Responders survived a median of 475 days (95% CI: 284 – 666 d), compared to 143 days (95% CI: 134 – 152 d) in nonresponders (P ⬍ .0001).

benefited from additional treatments. The changes in serum chemistry values and peripheral blood cell counts support a hypothesis that this unusual radiographic pattern represents an inflammatory response, tumor devascularization, and tumor lysis, but this remains to be confirmed. Postviral infusion biopsies of target tumors, limited by potential sampling error, showed only necrosis without cellular infiltration, and biopsies of adjacent parenchyma did not reveal cytopathic or vascular damage that would help explain the radiographic response pattern. The mechanisms and relative contributions of direct viral cytotoxicity, chemosensitization, cytokine induction, and immunomodulation to the efficacy of Onyx-015 remain an intriguing and complex issue. Comparatively, the survival statistics are more encouraging than the tumor size responses. This could be explained by the unusually slow regression of tumor sizes despite extensive necrosis within the lesions. Unfortunately, morphology of lesions in CT scans does not allow accurate quantification of necrosis and viability, and the total lesion size measured undoubtedly includes a large amount of necrotic debris. Why these lesions regress so slowly is unclear but may involve alterations in vascularity and increases in encapsulation. Only one patient underwent positron emission tomography and only one underwent

resection (both of which confirmed near-complete necrosis), and future investigations will need to include a more consistent method of quantification of tumor viability. Acknowledgments: The authors thank their coinvestigators, oncologists Evanthia Galanis, James Abbruzzese, Joseph Rubin, and David Kirn, for their collaboration. References 1. Panis Y, Rad AR, Boyer O, Houssin D, Salzmann JL, Klatzmann D. Gene therapy for liver tumors. Surg Oncol Clin N Am 1996; 5:461– 473. 2. Voss SD, Kruskal JB. Gene therapy: a primer for radiologists. Radiographics 1998; 18:1343–1372. 3. Oliff A, Gibbs JB, McCormick F. New molecular targets for cancer therapy. Sci Am 1996; 275:144 –149. 4. Wilson JM. Adenoviruses as gene-delivery vehicles. N Engl J Med 1996; 334: 1185–1187. 5. Miller HI. Gene therapy on trial. Science 2000; 287:591–592. 6. Kirn D. Replication-selective oncolytic adenoviruses: virotherapy aimed at genetic targets in cancer. Oncogene 2000; 19:6660 – 6669. 7. Galanis E, Vile R, Russell SJ. Delivery systems intended for in vivo gene therapy of cancer: targeting and replication competent viral vectors. Crit Rev Oncol Hematol 2001; 38:177–192. 8. Zwiebel JA. Cancer gene and oncolytic virus therapy. Semin Oncol 2001; 28:336 –343. 9. Bischoff JR, Kirn DH, Williams A, et al. An adenovirus mutant that replicates selectively in p53-deficient human tumor cells. Science 1996; 274:373–376.

•

289

10. Harris CC. Structure and function of the p53 tumor suppressor gene: clues for rational cancer therapeutic strategies. J Natl Cancer Inst 1996; 88:1442– 1455. 11. Heise C, Sampson-Johannes A, Williams A, McCormick F, Von Hoff DD, Kirn DH. ONYX-015, an E1B geneattenuated adenovirus, causes tumorspecific cytolysis and antitumoral efficacy that can be augmented by standard chemotherapeutic agents. Nat Med 1997; 3:639 – 645. 12. Reid T, Galanis E, Abbruzzese J, et al. Intra-arterial administration of a replication-selective adenovirus (dl1520) in patients with colorectal carcinoma metastatic to the liver: a phase I trial. Gene Ther 2001; 8:1618 –1626. 13. Moertel CG, Fleming TR, Macdonald JS, et al. Levamisole and fluorouracil for adjuvant therapy of resected colon carcinoma. N Engl J Med 1990; 322: 352–358. 14. Reid T, Galanis E, Abbruzzese J, Sze D, Wein L, Andrews J, Randlev B, Heise C, Uprichard M, Hatfield M, Romel L, Rubin J, and Kirn D. Hepatic arterial infusion of a replication-selective oncolytic adenovirus (dl1520): Phase II viral, immunologic, and clinical endpoints. Cancer Res 2002; 62:6070-6079. 15. Piedra PA, Poveda GA, Ramsey B, McCoy K, Hiatt PW. Incidence and prevalence of neutralizing antibodies to the common adenoviruses in children with cystic fibrosis: implication for gene therapy with adenovirus vectors. Pediatrics 1998; 101:1013–1019. 16. Letourneau JG, Thompson WM, Goldberg ME, Snover DC, Grage TB, Frick MP. Progressive CT appearance of hepatic metastases from colorectal carcinoma. Gastrointest Radiol 1988; 13: 145–151. 17. Havelaar IJ, Sugarbaker PH, Vermess M, Miller DL. Rate of growth of intraabdominal metastases from colorectal cancer. Cancer 1984; 54:163–171. 18. Mulvihill S, Warren R, Venook A, et al. Safety and feasibility of injection with an E1B-55 kDa gene-deleted, replication-selective adenovirus (ONYX-015) into primary carcinomas of the pancreas: a phase I trial. Gene Ther 2001; 8:308 –315. 19. Khuri FR, Nemunaitis J, Ganly I, et al. a controlled trial of intratumoral ONYX-015, a selectively-replicating adenovirus, in combination with cisplatin and 5-fluorouracil in patients with recurrent head and neck cancer. Nat Med 2000; 6:879 – 885. 20. Nemunaitis J, Khuri F, Ganly I, et al. Phase II trial of intratumoral administration of ONYX-015, a replication-selective adenovirus, in patients with re-

290

21.

22.

23.

24.

25.

•

Intraarterial Adenovirus for Metastatic Gastrointestinal Cancer

fractory head and neck cancer. J Clin Oncol 2001; 19:289 –298. Huard J, Lochmuller H, Acsadi G, Jani A, Massie B, Karpati G. The route of administration is a major determinant of the transduction efficiency of rat tissues by adenoviral recombinants. Gene Ther 1995; 2:107–115. Peeters MJ, Patijn GA, Lieber A, Meuse L, Kay MA. Adenovirus-mediated hepatic gene transfer in mice: comparison of intravascular and biliary administration. Hum Gene Ther 1996; 7:1693–1699. de Roos WK, Fallaux FJ, Marinelli AW, et al. Isolated-organ perfusion for local gene delivery: efficient adenovirusmediated gene transfer into the liver. Gene Ther 1997; 4:55– 62. Chia SH, Geller DA, Kibbe MR, et al. Adenovirus-mediated gene transfer to liver grafts: an improved method to maximize infectivity. Transplantation 1998; 66:1545–1551. Gerolami R, Cardoso J, Bralet MP, et al. Enhanced in vivo adenovirus-mediated gene transfer to rat hepatocarcino-

26.

27.

28.

29.

30.

mas by selective administration into the hepatic artery. Gene Ther 1998; 5:896 –904. Delman KA, Bennett JJ, Zager JS, et al. Effects of preexisting immunity on the response to herpes simplex-based oncolytic viral therapy. Hum Gene Ther 2000; 11:2465–2472. Kemeny N, Huang Y, Cohen AM, et al. Hepatic arterial infusion of chemotherapy after resection of hepatic metastases from colorectal cancer. N Engl J Med 1999; 341:2039 –2048. Ragnhammar P, Hafstrom L, Nygren P, Glimelius B. A systematic overview of chemotherapy effects in colorectal cancer. Acta Oncol 2001; 40:282– 308. Heise CC, Williams AM, Xue S, Propst M, Kirn DH. Intravenous administration of ONYX-015, a selectively replicating adenovirus, induces antitumoral efficacy. Cancer Res 1999; 59:2623–2628. Kuszyk BS, Corl FM, Franano FN, et al. Tumor transport physiology: implications for imaging and imaging-guided

31.

32.

33. 34.

35.

March 2003

JVIR

therapy. AJR Am J Roentgenol 2001; 177:747–753. Tomko RP, Xu R, Philipson L. HCAR and MCAR: the human and mouse cellular receptors for subgroup C adenoviruses and group B coxsackieviruses. Proc Natl Acad Sci U S A 1997; 94: 3352–3356. McCormick F. Interactions between adenovirus proteins and the p53 pathway: the development of ONYX-015. Semin Cancer Biol 2000; 10:453– 459. Wodarz D. Viruses as antitumor weapons: defining conditions for tumor remission. Cancer Res 2001; 61:3501–3507. Ries SJ, Brandts CH, Chung AS, et al. Loss of p14ARF in tumor cells facilitates replication of the adenovirus mutant dl1520 (ONYX-015). Nat Med 2000; 6:1128 –1133. Yang CT, You L, Uematsu K, Yeh CC, McCormick F, Jablons DM. p14(ARF) modulates the cytolytic effect of ONYX-015 in mesothelioma cells with wild-type p53. Cancer Res 2001; 61:5959 –5963.