pancreatic cancer hent1 expression and survival from ... · score for their cores was above the...

DOI:10.1093/jnci/djt347Advance Access publication December 3, 2013

JNCI | Article 1 of 10

© The Author 2013. Published by Oxford University Press. All rights reserved. For Permissions, please e-mail: [email protected].

jnci.oxfordjournals.org

Article

Pancreatic cancer heNt1 expression and Survival From Gemcitabine in Patients From the eSPAc-3 trialWilliam Greenhalf, Paula Ghaneh, John P. Neoptolemos, Daniel H. Palmer, Trevor F. Cox, Richard F. Lamb, Elizabeth Garner, Fiona Campbell, John R. Mackey, Eithne Costello, Malcolm J. Moore, Juan W. Valle, Alexander C. McDonald, Ross Carter, Niall C. Tebbutt, David Goldstein, Jennifer Shannon, Christos Dervenis, Bengt Glimelius, Mark Deakin, Richard M. Charnley, François Lacaine, Andrew G. Scarfe, Mark R. Middleton, Alan Anthoney, Christopher M. Halloran, Julia Mayerle, Attila Oláh, Richard Jackson, Charlotte L. Rawcliffe, Aldo Scarpa, Claudio Bassi, Markus W. Büchler; for the European Study Group for Pancreatic Cancer

Manuscript received June 17, 2013; revised October 24, 2013; accepted October 29, 2013.

Correspondence to: J. P. Neoptolemos, MD, Liverpool Cancer Trials Unit, Cancer Research UK Centre, University of Liverpool, 5th Fl UCD Bldg, Daulby St, Liverpool, L69 3GA, UK (e-mail: [email protected]).

Background Human equilibrative nucleoside transporter 1 (hENT1) levels in pancreatic adenocarcinoma may predict survival in patients who receive adjuvant gemcitabine after resection.

Methods Microarrays from 434 patients randomized to chemotherapy in the ESPAC-3 trial (plus controls from ESPAC-1/3) were stained with the 10D7G2 anti-hENT1 antibody. Patients were classified as having high hENT1 expression if the mean H score for their cores was above the overall median H score (48). High and low hENT1-expressing groups were compared using Kaplan–Meier curves, log-rank tests, and Cox proportional hazards models. All statistical tests were two-sided.

Results Three hundred eighty patients (87.6%) and 1808 cores were suitable and included in the final analysis. Median overall survival for gemcitabine-treated patients (n = 176) was 23.4 (95% confidence interval [CI] = 18.3 to 26.0) months vs 23.5 (95% CI = 19.8 to 27.3) months for 176 patients treated with 5-fluorouracil/folinic acid (χ2

1=0.24; P = .62). Median survival for patients treated with gemcitabine was 17.1 (95% CI = 14.3 to 23.8) months for those with low hENT1 expression vs 26.2 (95% CI = 21.2 to 31.4) months for those with high hENT1 expression (χ2

1= 9.87; P = .002). For the 5-fluorouracil group, median survival was 25.6 (95% CI = 20.1 to 27.9) and 21.9 (95% CI = 16.0 to 28.3) months for those with low and high hENT1 expression, respectively (χ2

1 = 0.83; P = .36). hENT1 levels were not predictive of survival for the 28 patients of the observation group (χ2

1 = 0.37; P = .54). Multivariable analysis confirmed hENT1 expression as a predictive marker in gemcitabine-treated (Wald χ2 = 9.16; P = .003) but not 5-fluorouracil–treated (Wald χ2 = 1.22; P = .27) patients.

Conclusions Subject to prospective validation, gemcitabine should not be used for patients with low tumor hENT1 expression.

JNCI J Natl Cancer Inst (2014) 106(1): djt347

Pancreatic ductal adenocarcinoma is a leading cause of cancer death and is characterized by early micrometastatic disease (1). most patients have advanced noncurable disease, but survival can be prolonged with mono or combination chemotherapy (2,3) using 5-fluorouracil (5Fu)–based regimens including FoLFIrINoX (4) or using gemcitabine monotherapy or couplets such as gemcit-abine with nab-paclitaxel (5). Surgical resection, which is possible in 15% to 20% of patients, can provide 5-year survival in only 10% of these (3). Although the greatly improved survival advantage for adjuvant chemotherapy is now well established (6–10), there is no randomized phase III clinical trial supporting evidence for the use of adjuvant chemoradiotherapy over systemic chemotherapy alone (3,11–13). Although the growing understanding of the molecular pathology of pancreatic cancer may well lead to more targeted ther-apies (14,15), there is now considerable opportunity for improved

outcomes based on current therapies through the stratified medi-cines approach using appropriate response biomarkers (3,15).

A promising biomarker is the human equilibrative nucleoside transporter 1 (heNt1), which permits the bidirectional passage into cells of pyrimidine nucleosides such as gemcitabine capecit-abine and 5Fu (16,17). the original clinical study, which was undertaken using stored tissue from 21 patients with advanced pancreatic cancer treated with gemcitabine, showed that cancer tissue from patients with greater heNt1 expression by immu-nohistochemistry had a longer survival (17). Subsequent studies have provided encouraging but not convincing evidence because they have been mostly retrospective and nonrandomized series (18,20–25), with mixed advanced and adjuvant cases (18), based on neoadjuvant (24,25) or adjuvant chemoradiotherapy (18–22) or using combination regimens such as gemcitabine with S-1, which

mailto:[email protected]?subject=

Vol. 106, Issue 1 | djt347 | January 1, 20142 of 10 Article | JNCI

is a 5Fu-prodrug (23). None of these latter studies truly tested gemcitabine monotherapy in a defined group of patients (18–25). moreover patients with a better performance status are more likely to receive gemcitabine treatment, whereas those with poor perfor-mance are more likely to receive no or only limited treatment. thus it is not possible to discriminate between a therapeutic predictive effect and a disease prognostic effect for heNt1 expression.

there is, however, considerable opportunity to minimize biases in biomarker development for pancreatic cancer by using clini-cal data and cancer tissue collected from prospective, randomized controlled trials conducted by the european Study Group for Pancreatic Cancer (eSPAC) (trial registration clinicaltrials.gov: NCt00058201). eSPAC-1 showed that adjuvant chemotherapy using 5Fu/folinic acid improved overall survival but that adjuvant chemoradiotherapy did not (6,7). eSPAC-3(v2) demonstrated that adjuvant gemcitabine was not superior to 5Fu/folinic acid (9). therefore our main aim was to determine whether tumor heNt1 levels would help predict overall survival for patients receiving gemcitabine compared with patients receiving 5Fu.

MethodsStudy Designthe translational eSPAC-t studies received ethical committee approval for characterization of tumor markers for chemother-apy from the Liverpool (Adult) research ethics Committee (07/h1005/87). use of Good Clinical Practice standard operating proce-dures (26) ensures a full audit trail and prevents access to outcome data by pathologists and laboratory researchers. After resection for pancre-atic ductal adenocarcinoma, patients in the eSPAC-1 (6,7) 2 × 2 facto-rial trial were randomized either to receive chemoradiotherapy or not and also to receive chemotherapy using 5Fu and folinic acid or not. In the eSPAC-3 study (9), patients were randomized to observation or to receive either 5Fu with folinic acid, as in the eSPAC-1 trial, or gem-citabine. once the definitive results of eSPAC-1 were known (7), the observation arm was removed, and the trial was renamed eSPAC-3(v2) (9). Both studies were analyzed on an intention-to-treat basis, but for the eSPAC-t study, patients in the treatment arms were selected for inclusion only if treatment was actually received. All patients provided written informed consent. this study was conducted and reported in accordance with the remArK criteria (27).

Tissue Microarray Manufacturethe arrays were manufactured using standard operating procedures conducted to Laboratory Good Clinical Practice. the treatment arrays contained cores from patients entirely from eSPAC-3(v2) randomized to either 5Fu/folinic acid or to gemcitabine. the obser-vation array contained cores from both eSPAC-1 and eSPAC-3 because of the paucity of samples from patients truly randomized to observation. Although fixation protocols for formalin-fixed paraffin-embedded tissue blocks varied across centers, they were standardized for each center with no evidence for center- or country-specific bias in heNt1 staining (see Supplementary table 1, available online). Cores were taken from tumor regions identified by an experienced pancreatic pathologist (F. Campbell) using hematoxylin and eosin–stained sections. tissue microarrays were prepared with two cores from each block, with four to eight cores arrayed for each patient.

each of the treatment eSPAC-3(v2) arrays had two cores from 88 patients, whereas the observation array had four cores from 31 patients. For all arrays, control cores, comprising three cores each of colon, kidney, liver, normal pancreas, and chronic pancreatitis, were arranged in a fence around the test samples.

Antibody Validationthe mouse monoclonal anti-heNt1 antibody 10D7G2 was used (17,28). Western blotting revealed a series of bands between 30 KDa and 70K Da in the pancreatic cancer cell lines Panc-1, Suit-2, miaPaCa, and BxPC3. Immunoblotting confirmed that 10D7G2 detected a full heNt1 sequence GSt-tagged recom-binant protein (SLC29A1 human recombinant protein; Abnova taiwan) at concentrations of 3 0ng to 200 ng per lane. mass spectrometry (triple toF-5600; AB Sciex, Warrington, uK) of 10D7G2 immunoprecipitate from polyacrylamide gel bands pre-pared from Panc-1 and miaPaCa cells revealed a single heNt1 peptide (AGKeeSGVSVSNSQPtNeShSIK) in the 50 KDa band, which was not seen with the control immunoprecipitate (beads but no antibody). Clean-up of the antibody was required using preabsorption to reduce background and improve specific-ity before Western blotting. Preabsorption was performed using 5 mL of 1:500 dilution of 10D7G2 in 3% bovine serum albumin in phosphate-buffered saline/0.1% tween. A 10% tris glycine gel was run with 20 µg total protein per well from BxPC3, Suit2 Panc1, miaPaCa, and heLa cell lines and transferred to a polyvinylidene fluoride membrane. the membrane was blocked in 3% bovine serum albumin diluted in phosphate-buffered saline/0.1% tween. Immunoblotting was performed at room temperature with agita-tion for 1 hour. this removed some of the nonspecific binding of the antibody. the antibody solution was then decanted and used to immunoblot polyvinylidene fluoride membranes containing lysate from knockdown experiments and the immunoprecipitation lysate.

heLa wild-type cells, heLa with heNt1 knockdown, and heLa cells engineered to overexpress heNt1 were grown in 2 × t150 flasks to confluency. the cells were pelleted by centrifugation and fixed in 10% neutral buffered formalin overnight. Cells were then mixed in a 1:1 ratio with 4% agar and set in a mold. the resulting cell pellet was formalin fixed and embedded in paraffin. Five micrometer sections were prepared on Superfrost Plus slides. A 1:2 dilution of 10D7G2 antibody in 0.05mol/L tris-hCl buffer ph 7.2 was used for immu-nocytochemistry. heNt1 expression in heLa cells showed strong membrane staining along with some cytoplasmic and nuclear staining. heNt1 knockdown by small interfering rNA reduced membrane and cytoplasmic staining in heLa cells, whereas heNt1-overexpressing heLa cells resulted in increased membrane and cytoplasmic staining.

Immunohistochemistrythe heNt1 10D7G2 antibody was diluted 1:2 in 0.05 mol/L tris-hCl buffer ph 7.2 for immunohistochemistry on 5-µm sections of the tissue microarrays. After deparaffinization and hydration of sections, antigens were retrieved by heat treatment in citrate buffer ph 6.0. Sections were treated with peroxidase block. Slides were incubated with the primary antibody in tris buffer (as above) at 4 oC overnight. Immunostaining was performed by the avidin-biotin peroxidase com-plex technique (enVision System; Dako, ely, uK). Negative controls were incubated with buffer alone in place of primary antibody.

http://jnci.oxfordjournals.org/lookup/suppl/doi:10.1093/jnci/djt347/-/DC1

JNCI | Article 3 of 10jnci.oxfordjournals.org

Cores were scored by a specialist pancreas histopathologist (F. Campbell) assisted by a technician (e. Garner) to record the find-ings for each core. F. Campbell was blinded to treatment group when scoring. Both F. Campbell and e. Garner were blinded to patient outcomes throughout the study. the intensity of cytoplas-mic and membrane staining in tumor cells was ranked from 1 to 3. Lymphocyte staining was taken as an internal positive control (9). h scores were produced for each core by multiplying inten-sity score by the percentage of tumor cells stained, and a mean h score was calculated for each patient based on four to eight cores (Supplementary Figure 1, available online).

Statistical Analysisthe largest previous study had reported an unadjusted hazard ratio (hr) in favor of heNt1 levels predicting response to adjuvant therapy (albeit a mix of adjuvant chemotherapy and chemoradio-therapy regimens that included gemcitabine at some point) of 0.43 (95% confidence interval [CI] = 0.29 to 0.63; n = 243) (21). A more conservative power calculation was used in our study based on 175 patients with 137 deaths within each treatment arm, which, assum-ing a split ratio of 1:1 of high to low heNt1 expression, would detect a hazard ratio of 0.62 between the two levels of expression, for a statistical significance level of .05 at 80% power.

All analyses were carried out using SAS version 9.2 software (SAS Institute, Cary, NC). Survival from date of randomization

was analyzed by the method of Kaplan–meier (29), with differences between groups assessed using the mantel–Cox log rank test with multiple comparisons adjusted by the method of tukey–Kramer. recurrence-free survival was time to recurrence as assessed by com-puterised tomography or time to death. univariate and multivariable analyses were carried out using Cox proportional hazards. the check-ing of proportionality and model fit was carried out using plots of log(-log(survival)) against log(survival time) and martingale and deviance residual plots.

A log transformation was also made on the h scores because of the skewed data distribution and where zero values were given a notional value of 1 to allow the use of logarithms. Variables included in the analyses were sex and country but not ethnicity because this information had not been collected. All statistical tests were two-sided

resultPatients and Tissue Samplestwelve tissue microarrays representing 202 gemcitabine-treated patients, 201 patients treated with 5Fu/folinic acid, and 31 true resection only patients, were chosen for this study (Figure 1). three hundred eighty patients (87.6%) and 1808 cores were included in the final analysis, and the demographic, surgery, and pathology fea-tures of the patients included are shown in table 1.

Figure 1. CONSORT diagram. 5FU/FA = 5-fluorouracil /folinic acid; FFPE = formalin-fixed paraffin-embedded; PDAC = pancreatic ductal adenocar-cinoma; TMA = tissue microarray.



Table 1. Demographic, surgery, and pathology features of the patients scored for hENT1*

Demographics and characteristics

Chemotherapy

Observation (n = 28)

Total (n = 380)

5-Fluorouracil/ folinic acid (n = 176)

Gemcitabine (n = 176)

Age, median (IQR), y 63 (57–70) 65 (57–72) 54 (52–69) 64 (57–70)Sex, No. (%) Female 77 (43.8) 70 (39.8) 7 (25.0) 154 (40.5) Male 99 (56.3) 106 (60.2) 21 (75.0) 226 (59.5)Baseline performance score, No. (%) 0 61 (34.7) 56 (31.8) 3 (27.3) 120 (33.1) 1 95 (54.0) 100 (56.8) 6 (54.5) 201 (55.4) 2 20 (11.4) 20 (11.4) 2 (18.2) 42 (11.6)Diabetic, No. (%) No 134 (80.2) 132 (79.0) 21 (77.8) 287 (79.5) NIDDM 13 (7.8) 13 (7.8) 3 (11.1) 29 (8.0) IDDM 20 (12.0) 22 (13.2) 3 (11.1) 45 (12.5)Smoking status, No. (%) Never 70 (44.0) 65 (41.9) 9 (34.6) 144 (42.4) Past 58 (36.5) 64 (41.3) 12 (46.2) 134 (39.4) Present 31 (19.5) 26 (16.8) 5 (19.2) 62 (18.2)Postoperative complications, No. (%) No 136 (79.5) 127 (75.1) 21 (80.8) 284 (78.0) Yes 35 (20.5) 42 (24.9) 5 (19.2) 82 (22.0)Hospital stay No. 159 158 10 327 Median (IQR) days 13 (10–16) 13 (10–17) 14 (12–18) 13 (10–17)Postoperative CA 19-9 No. 129 125 10 264 Median (IQR) KU/l 35 (13–80) 23 (11–48) 8 (2–39) 28 (11–68)Surgery to randomization Median (IQR) days 49 (35–61) 50 (38–62) 35 (25–52) 49 (36–61)Surgery, No. (%) Whipples resection 88 (54.0) 83 (52.8) 20 (83.3) 191 (55.5) Pylorus preserving 66 (40.5) 70 (44.6) 3 (12.5) 139 (40.4) Total pancreatectomy 9 (5.5) 4 (2.5) 1 (4.2) 14 (4.1)Extent of resection, No. (%) Standard 133 (80.1) 118 (70.7) 8 (88.8) 259 (75.7) Radical resection 22 (13.3) 26 (15.6) 1(11.2) 49 (14.3) Extended radical 11 (6.6) 23 (13.8) 0 (0.0) 34 (9.9)Maximum tumor diameter, mm Median (IQR) 30 (22–35) 30 (25–40) 30 (20–37) 30 (22–37)Tumor grade, No. (%) Well 13 (7.5) 15 (8.6) 4 (14.3) 32 (8.5) Moderately 118 (68.2) 114 (65.5) 17 (60.7) 249 (66.4) Poorly 42 (24.3) 45 (25.9) 7 (25.0) 94 (25.1)Lymph node invasion, No. (%) Negative 39 (22.2) 34 (19.3) 5 (17.9) 177 (21)

78 (20.5) Positive 137 (77.8) 142 (80.7) 23 (82.1) 251 (79)

302 (79.5)Resection margins, No. (%) Negative 97 (55.1) 103 (58.5) 19 (67.9) 219 (57.6) Positive 79 (44.9) 73 (41.5) 9 (32.1) 161 (42.4)Local Invasion, No. (%) No 95 (55.6) 88 (52.4) 15 (55.6) 198 (54.1) Yes 76 (44.4) 80 (47.6) 12 (44.4) 168 (45.9)Tumor stage, No. (%) I 13 (7.4) 9 (5.2) 0 (0.0) 22 (6.1) II 46 (26.3) 45 (25.9) 2(20.0) 93 (25.9) III 109 (62.3) 116 (66.7) 8 (80.0) 233 (64.9) IVa 7 (4.0) 4 (2.3) 0 (0.0) 11 (3.1)

* IDDM = insulin-dependent diabetes mellitus; IQR = interquartile range; NIDDM = noninsulin-dependent diabetes mellitus.


hENT1 Stainingthe overall h score median was 48, interquartile range (IQr) was 10 to 80, and range was 0 to 210. the median h score was 42 (IQr = 3–69) for the 5Fu/folinic acid group, 50 (IQr = 24–93) for the gemcitabine group, and 86 (IQr = 40–101) for the observation group. using the overall median h score of 48 as the threshold, in the gemcitabine group (n = 176), there were 146 (83.0%) deaths with 74 (50.7%) above the threshold. In the 5Fu/folinic acid group (n = 176), there were 139 (79.0%) deaths with 60 (43.2%) above the threshold. In the observation group (n = 28), there were 23 (82.1%) deaths with 18 (78.3%) above the threshold. Contingency testing of heNt1 levels with clinical and tumor characteristics (table 2) did not reveal any statistically significant associations except that a younger age might be associated with higher heNt1 tumor expression (crude P = .02), although this was not statistically sig-nificant after correction for multiple testing.

Overall Survival Analysisthe median overall survival for the gemcitabine-treated patients was 23.4 (95% CI = 18.3 to 26.0) months compared with 23.5 (95% CI = 19.8 to 27.3) months for the patients treated with 5Fu/folinic acid (χ2

1= 0.241; P = .62). the median survival for the patients treated with gemcitabine with low heNt1 levels was 17.1 (95%

CI = 14.3 to 23.8) months compared with 26.2 (95% CI = 21.2 to 31.4) months for those with high heNt1 levels (χ2

1 = 9.87; P = .002), with hazard ratio of 0.60 (95% CI = 0.43 to 0.83) for high expressers compared with low expressers (Figure 2). thus our primary null-hypothesis was rejected, indicating a difference of 9.1 months in median survival for high and low heNt1 expressers in gemcitabine-treated patients.

the median survival in patients treated with 5Fu/folinic acid was 25.6 (95% CI = 20.1 to 27.9) months for those with low heNt1 levels and 21.9 (95% CI = 16.0 to 28.3) months for those with high heNt1 levels (χ2

1 = 0.83; P = .36), with a hazard ratio of 1.17 (95% CI = 0.84 to 1.64).

Survival for the pure observation patients did not differ by heNt1 expression regardless of whether the overall median h score (χ2

1 = 0.37; P = .54) or the median h score just for patients in the observation group was used as the threshold (Supplementary Figure 2, available online).

A further analysis was also carried out using a Cox proportional hazards model with log(heNt1) as a covariable for each chemo-therapy arm. For the gemcitabine arm, log(heNt1) was a statisti-cally significant covariable (P = .03), with hazard ratio of 0.80 per 10 units of heNt1. For the 5Fu arm, log(heNt1) was not a sta-tistically significant covariable (P = .90).

Table 2. Relationship between hENT1 level and patient or tumor characteristics*

Characteristic NumberhENT1 H score, median (IQR)

Low hENT1 (<48), No.

High hENT1 (≥48), No. P

Resection margin Negative Positive

219161

47 (3–83)50 (10–80)

11478

10583

.53†

Lymph node status Negative Positive

78302

47 (17–88)49 (5–80)

42150

36152

.53†

Stage 1 2 3 4

2293

23311

39 (23–60)50 (10–83)48 (10–75)17 (0–50)

1444

1208

849

1133

.27‡

Tumor grade Well Moderate Poor

3224994

58 (19–95)50 (10–83)45 (0–73)

1412152

1812842

.42‡

Local invasion No Yes

198168

50 (23–80)47 (0–79)

9688

10280

.47†

Maximum tumor diameter <30 mm ≥30 mm

178202

50 (23–85)45 (0–80)

83109

9593

.18†

Diabetes No IDDM NIDDM

2874529

49 (3–79)47 (0–85)47 (17–85)

1432216

1442313

.85‡

Sex Male Female

226154

48 (5–88)48 (12–78)

11577

11177

.92†

Age, y <64 ≥64

189191

50 (13–85)45 (5–76)

84108

10583

.02†

* IDDM = insulin-dependent diabetes mellitus; IQR = interquartile range; NIDDM = noninsulin-dependent diabetes mellitus.

† Fisher exact test, two-sided value.

‡ χ2 test, two-sided value.




univariate analysis of the resection margin status, lymph node status, tumor stage, local invasion, and maximum tumor diameter all showed the expected statistically significant association with survival. Differences in survival with local invasion, diabetes melli-tus, sex, and age did not reach statistical significance. tumor grade, smoking, and heNt1 expression were only shown to be statisti-cally significantly associated with survival in patients randomized to gemcitabine (table 3). multivariable analysis of survival to allow for resection margin status, lymph node status, tumor grade, sex, and smoking (table 4) was statistically significant for the heNt1 expression in the gemcitabine group (Wald χ2 = 9.16; P = .003) but not for the 5Fu/folinic acid group (Wald χ2 = 1.22; P = .27).

Pairwise comparisons of all variations of heNt1 level and treatment arms showed that only in the gemcitabine arm was high

heNt1 statistically significantly associated with survival (χ2 = 8.62; P = .02) (Supplementary table 4, available online). A series of h score sensitivity analyses was undertaken, including removing the middle 15 and 30 percentiles without affecting the principal conclu-sion (Supplementary Figures 3 and 4, available online). Similarlyl taking only the top and bottom 33% of each treatment group (Supplementary Figure 5, available online) and normalization of the scores for each array (Supplementary Figure 6, available online) gave the same trends.

Recurrence-Free Survivalrecurrence-free survival curves mirrored the findings for overall sur-vival (Supplementary Figure 7, available online). median recurrence-free survival for gemcitabine-treated patients was 11.1 (95% CI = 8.8

Figure 2. Survival curves split by treatment group (5-fluorouracil /folinic acid [5FU] or gemcitabine [GEM]) and hENT1 expression levels (high = mean H score ≥48; low = mean H score <48). All groups and the number of at risk individuals are shown in the top graph, and below the data are sepa-rated out for ease of viewing of pairwise comparisons. All statistical tests were log-rank analyses using two-sided χ2 tests. CI = confidence interval.








to 13.8) months for those with low heNt1 levels and 14.3 (95% CI = 11.6 to 16.6) months for those with high heNt1 levels. For the gemcitabine-treated patients, there was a statistically significant difference in recurrence-free survival between high and low heNt1 expressers (χ2 = 7.45′ P = .006). median recurrence-free survival for 5Fu-treated patients was 12.5 months (95% CI = 9.7 to 17.8) months for those with low heNt1 expression and 12.8 (95% CI = 10.3 to 15.9) months for those with high heNt1 expression (χ2 = 1.43; P = .23).

Discussionthis is the first study to test heNt1 expression levels in an unbi-ased group of patients that were given gemcitabine monotherapy in the identical clinical setting by using tissues and clinical data from the treatment arms used in the same randomized trial. thus we only included treated patients from the adjuvant eSPAC3(v2) trial (9). Because we wished also to explore analysis in a pure con-trol group specifically randomized to no adjuvant treatment, this

Table 3. Univariate analysis of survival factors*

Characteristic

HR (95% CI)

Total

Chemotherapy

5-Fluorouracil /folinic acid Gemcitabine

Resection margin Negative Positive

n = 176 n = 176 n = 3521.00 (referent) 1.00 (referent) 1.00 (referent)

1.66 (1.18 to 2.32) 1.34 (0.97 to 1.87) 1.50 (1.18 to 1.89)Wald χ2 = 8.63, P = .003 Wald χ2 = 3.09, P = .08 Wald χ2 = 11.32, P = .001

Lymph node status Negative Positive


2.15 (1.38 to 3.34) 1.88 (1.19 to 2.95) 2.04 (1.49 to 2.80)Wald χ2 = 11.39, P = .001 Wald χ2 = 7.34, P = .007 Wald χ2 = 19.35, P < .001

Stage 1 2 3 4


1.45 (0.64 to 3.27) 1.15 (0.48 to 2.74) 1.33 (0.74 to 2.40)2.04 (0.94 to 4.42) 1.70 (0.75 to 3.88) 1.92 (1.09 to 3.38)1.84 (0.62–5.48) 0.64 (0.13 to 3.16) 1.27 (0.53 to 3.06)

Wald χ2 = 5.43, P = .14 Wald χ2 = 6.45, P = .09 Wald χ2 = 11.35, P = .01Tumor grade Well Moderately Poorly


0.83 (0.44 to 1.55) 1.01 (0.55 to 1.85) 0.93 (0.60 to 1.43)0.85 (0.43 to 1.71) 1.67 (0.87 to 3.19) 1.20 (0.75 to 1.92)

Wald χ2 = 0.35, P = 84 Wald χ2 = 7.16, P = .03 Wald χ2 = 3.39, P = .18Local invasion No Yes



Maximum tumor diameter <30mm ≥30mm



Diabetes mellitus No Yes



Sex Male Female

n = 176 n = 176 = 3521.00 (referent) 1.00 (referent) 1.00 (referent)


Age, y ≥64 <64



Smoking Never Ex Current


1.22 (0.82 to 1.81) 1.16 (0.79 to 1.70) 1.19 (0.90 to 1.56)1.32 (0.83 to 2.11) 1.70 (1.05 to 2.77) 1.46 (1.04 to 2.04)

Wald χ2 = 1.72, P = .42 Wald χ2 = 4.60, P = .10 Wald χ2 = 5.04, P = .08hENT1 expression Low High



* All statistical tests were two-sided. CI = confidence interval; HR = hazard ratio.


meant combining the observation arms from the eSPAC-1 (6,7) and eSPAC-3 (9) trials. Although the resultant number of cases investigated was small, there was no evidence to support heNt1 expression levels per se as indicative of overall survival. taken together, the results of this study have provided powerful evidence that adjuvant gemcitabine should not be used after resection for patients with low tumor heNt1 expression, the alternative being 5Fu-based regimens, and by implication vice versa.

the bias of nonrandomized studies makes it impossible to separate a predictive therapeutic-specific effect from a disease prognostic-specific effect. As well as using patients from retro-spective uncontrolled cohorts, methodological considerations may have contributed to the variability of previously published studies. Differences in heNt1 rNA transcript levels between studies can be difficult to interpret given the variances in tissue collection, storage, retrieval, and analytical techniques; in one study showed that higher heNt1 transcript levels were associated with longer survival (18), but in another study heNt1 transcript levels appeared to be prog-nostic irrespective of gemcitabine use (20). three studies of adjuvant gemcitabine and chemoradiation that showed a survival benefit for higher heNt1 levels used the 10D7G2 monoclonal murine anti-body (19,20,21), whereas the study that showed an opposite benefit with low heNt1 expression used a Sigma-Aldrich polyclonal anti-body (23) and the adjuvant study of gemcitabine and S-1 used an Abnova polyclonal rabbit antibody (22). on the other hand these dif-ferences are just as likely to be due to stage mix and patient selection.

It is not entirely certain as to how heNt1 expression levels in resected specimens may relate to the residual micrometastases respon-sible for recurrence after treatment (1). one potential explanation for the interaction between heNt1 expression levels and benefit from specific therapies is that heNt1 has a far greater affinity for nucleo-sides such as gemcitabine than for nucleobases such as 5Fu (30,31). Also other proteins, including the human organic anion transporter 2 (32) or heNt2 (33), may more readily mediate 5Fu uptake.

the one previous randomized study in which heNt1 expression was explored was in a small subset of the rtoG-9704 trial, essen-tially an adjuvant chemoradiotherapy trial, with patients randomized to short courses of 5Fu or gemcitabine both before and after chem-oradiotherapy and using 5Fu as the radiosensitizer in both arms (19). A high level of heNt1 expression was associated with a sur-vival benefit compared with low heNt1 levels in the gemcitabine 5Fu-chemoradiotherapy–treated group. Interpretation is under-mined, however, because the survival benefit of adjuvant chemora-diotherapy remains unproven in pancreatic cancer (3,11–13).

the recently reported randomized phase II advanced pancreatic cancer LeAP trial of a membrane permeable form of gemcitabine (Co-101) failed to meet its primary endpoint (34). In this trial, no survival advantage in gemcitabine-treated patients was seen for high heNt1 expression determined by immunohistochemistry using the SP120 rabbit monoclonal antibody. this SP120 antibody was only previously tested in a small subset of the 5Fu-based chemoradiation rtoG-9704 trial using primary tumor tissue, whereas in the LeAP trial heNt1 expression appears to have been largely determined in metastatic tissue. It remains uncertain whether heNt1 expres-sion levels can provide similar predictive therapeutic information between adjuvant gemcitabine with 5Fu-chemoradiotherapy in the post-resection setting on the one hand and palliative gemcitabine in the advanced and metastatic setting on the other hand. the per-formance of the SP120 antibody might also be in question, but it is now not possible to compare this rabbit antibody with the 10D7G2 murine antibody because the former is no longer available. overall then it is unclear whether the SP120 antibody was adequately vali-dated by examining heNt1 expression levels in metastatic pan-creas cancer tissues in patients who had received gemcitabine as the only therapy before testing in the LeAP trial. In contrast heNt1 expression using the 10D7G2 antibody has been tested in patients who have received gemcitabine monotherapy in both the advanced (17) and now the adjuvant setting with consistent results. our study

Table 4. Multiple regression analysis of survival factors by multivariable analysis*

Variable

5-Fluorouracil/folinic acid (n = 156) Gemcitabine (n = 153)

HR (95% CI) Wald χ2 P HR (95% CI) Wald χ2 P

Resection margin Negative 1.00 (referent) 6.48 .01 1.00 (referent) 2.06 .15 Positive 1.71 (1.13 to 2.57) 1.33 (0.90 to 1.96)Lymph node status Negative 1.00 (referent) 9.03 .003 1.00 (referent) 6.66 .01 Positive 2.17 (1.31 to 3.60) 1.97 (1.18 to 3.28)Tumor grade Well 1.00 (referent) 2.44 .30 1.00 (referent) 11.21 .004 Moderately 0.57 (0.27 to 1.17) 0.72 (0.36 to 1.43) Poorly 0.56 (0.25 to 1.24) 1.52 (0.71 to 3.23)Sex Male 1.0 (referent) 0.03 .86 1.00 (referent) 3.29 .07 Female 0.97 (0.65 to 1.43) 1.42 (0.97 to 2.08)Smoking status Never 1.00 (referent) 1.22 .55 1.00 (referent) 6.17 .046 Ex 1.20 (0.77 to 1.85) 1.20 (0.78 to 1.86) Present 1.30 (0.79 to 2.15) 1.98 (1.15 to 3.40)hENT1 expression Low 1.00 (referent) 1.22 .27 1.00 (referent) 9.16 .003 High 1.24 (0.85 to 1.83) 0.55 (0.37 to 0.81)

* All P values were calculated using a two-sided χ2 test based on the Wald χ2 test statistic. CI = confidence interval; HR = hazard ratio.


was not without limitations because patients were not treated with gemcitabine or 5Fu based on heNt1 expression. the validity of selecting patients for gemcitabine- and nongemcitabine-based regi-mens will need to be confirmed by prospective studies.

Nevertheless our study paves the way for the introduction of heNt1 assessment for a stratified medicines approach in pancreatic cancer, not only in the adjuvant setting but also potentially in the advanced setting.

references 1. tuveson DA, Neoptolemos JP. understanding metastasis in pancreatic

cancer: a call for new clinical approaches. Cell. 2012;148(1–2):21–23. 2. tempero mA, Arnoletti JP, Behrman SW, et al. Pancreatic adenocarci-

noma, version 2.2012: featured updates to the NCCN Guidelines. J Natl Compr Canc Netw. 2012;10(6):703–713.

3. Van Laethem JL, Verslype C, Iovanna JL, et al. New strategies and designs in pancreatic cancer research: consensus guidelines report from a european expert panel. Ann Oncol. 2012;23(3):570–576.

4. Conroy t, Desseigne F, ychou m, et al. FoLFIrINoX versus gemcitabine for metastatic pancreatic cancer. N Engl J Med. 2011;364(19):1817–1825.

5. Von hoff DD et al. results of a randomized phase III trial (mPACt) of weekly nab-paclitaxel plus gemcitabine versus gemcitabine alone for patients with metastatic adenocarcinoma of the pancreas with Pet and CA19-9 correlates. J Clin Oncol. 2013;13(Suppl): abstract 4005.

6. Neoptolemos JP, Dunn JA, moffitt DD, et al. eSPAC-1: a european, rand-omized controlled study of adjuvant chemoradiation and chemotherapy in resectable pancreatic cancer. Lancet. 2001;358(9293):1576–1585.

7. Neoptolemos JP, Stocken DD, Friess h, et al. A randomized trial of chem-oradiotherapy and chemotherapy after resection of pancreatic cancer. N Engl J Med. 2004;350(12):1200–1210.

8. oettle h, Post S, Neuhaus P, et al. Adjuvant chemotherapy with gemcitabine vs observation in patients undergoing curative-intent resection of pancre-atic cancer: a randomized controlled trial. JAMA. 2007;297(3):267–277.

9. Neoptolemos JP, Stocken DD, Bassi C, et al. Adjuvant chemother-apy with fluorouracil plus folinic acid vs gemcitabine following pancreatic cancer resection: a randomized controlled trial. JAMA. 2010;304(10):1073–1081.

10. Fukutomi A, uesaka K, Boku N, et al. JASPAC 01: randomized phase III trial of adjuvant chemotherapy with gemcitabine versus S-1 for patients with resected pancreatic cancer. .J Clin Oncol. 2013;31(Suppl): abstract 4008.

11. Stocken D, Büchler m, Dervenis C, et al. meta-analysis of randomised adju-vant therapy trials for pancreatic cancer. Br J Cancer. 2005;92(8):1372–1381.

12. regine WF, Winter KA, Abrams rA, et al. Fluorouracil vs gemcitabine chemotherapy before and after fluorouracil-based chemoradiation follow-ing resection of pancreatic adenocarcinoma: a randomized controlled trial. JAMA. 2008;299(9):1019–1026.

13. twombly r. Adjuvant chemoradiation for pancreatic cancer: few good data, much debate. J Natl Cancer Inst. 2008;100(23):1670–1671.

14. Li D, Abbruzzese JL. New strategies in pancreatic cancer: emerging epide-miologic and therapeutic concepts. Clin Cancer Res. 2010;16(17):4313–4318.

15. Costello e, Greenhalf W, Neoptolemos JP. New biomarkers and targets in pancreatic cancer and their application to treatment. Nat Rev Gastroenterol Hepatol. 2012;9(8):435–444.

16. yao Sy, Ng Am, Cass Ce, et al. Nucleobase transport by human equilibrative nucleoside transporter 1 (heNt1). J Biol Chem. 2011;286(37):32552–32562.

17. Spratlin J, Sangha r, Glubrecht D, et al. the absence of human equili-brative nucleoside transporter 1 is associated with reduced survival in patients with gemcitabine-treated pancreas adenocarcinoma. Clin Cancer Res. 2004;10(20):6956–6961.

18. Giovannetti e, Del tacca m, mey V, et al. transcription analysis of human equilibrative nucleoside transporter-1 predicts survival in pancreas cancer patients treated with gemcitabine. Cancer Res. 2006;66(7):3928–3935.

19. Farrell JJ, elsaleh h, Garcia m, et al. human equilibrative nucleoside transporter 1 levels predict response to gemcitabine in patients with pan-creatic cancer. Gastroenterology. 2009;136(1):187–195.

20. Kim r, tan A, Lai KK, et al. Prognostic roles of human equilibrative trans-porter 1 (heNt-1) and ribonucleoside reductase subunit m1 (rrm1) in resected pancreatic cancer. Cancer. 2011;117(14):3126–3134.

21. maréchal r, Bachet JB, mackey Jr, et al. Levels of gemcitabine trans-port and metabolism proteins predict survival times of patients treated with gemcitabine for pancreatic adenocarcinoma. Gastroenterology. 2012;143(3):664–674.

22. Fisher SB, Patel Sh, Bagci P, et al. An analysis of human equilibrative nucleoside transporter-1, ribonucleoside reductase subunit m1, ribonu-cleoside reductase subunit m2, and excision repair cross-complementing gene-1 expression in patients with resected pancreas adenocarcinoma: Implications for adjuvant treatment. Cancer. 2013;119(2):445–453.

23. Nakagawa N, murakami y, uemura K, et al. Combined analysis of intratu-moral human equilibrative nucleoside transporter 1 (heNt1) and ribonu-cleotide reductase regulatory subunit m1 (rrm1) expression is a powerful predictor of survival in patients with pancreatic carcinoma treated with adjuvant gemcitabine-based chemotherapy after operative resection. Surgery. 2013;153(4):565–575.

24. murata y, hamada t, Kishiwada m, et al. human equilibrative nucleoside transporter 1 expression is a strong independent prognostic factor in uICC t3-t4 pancreatic cancer patients treated with preoperative gemcitabine-based chemoradiotherapy. J Hepatobiliary Pancreat Sci. 2012;19(4):413–425.

25. Kawada N, uehara h, Katayama K, et al. human equilibrative nucleo-side transporter 1 level does not predict prognosis in pancreatic cancer patients treated with neoadjuvant chemoradiation including gemcitabine. J Hepatobiliary Pancreat Sci. 2012;19(6):717–722.

26. Sarzotti-Kelsoe m, Cox J, Cleland N, et al.: evaluation and recommenda-tions on good clinical laboratory practice guidelines for phase I-III clinical trials. PLoS Med. 2009;6(5):e1000067.

27. mcShane Lm, Altman DG, Sauerbrei W, et al. reporting recommenda-tions for tumor marker prognostic studies (remArK). J Natl Cancer Inst. 2005;97(16):1180–1184.

28. Jennings LL, hao C, Cabrita mA, et al. Distinct regional distribu-tion of human equilibrative nucleoside transporter proteins 1 and 2 (heNt1 and heNt2) in the central nervous system. Neuropharmacology. 2001;40(5):722–731.

29. Kaplan e, meier P. Non-parametric estimation from incomplete observa-tions. J Amer Stat Assoc. 1958;49(268):457–481.

30. Zhang J, Visser F, King Km, et al. the role of nucleoside transporters in cancer chemotherapy with nucleoside drugs. Cancer Metastasis Rev. 2007;26:85–110.

31. Ward JL, Sherali A, mo ZP, et al. Kinetic and pharmacological properties of cloned human equilibrative nucleoside transporters, eNt1 and eNt2, stably expressed in nucleoside transporter-deficient PK15 cells. ent2 exhibits a low affinity for guanosine and cytidine but a high affinity for inosine. J Biol Chem. 2000;275(12):8375–8381.

32. Kobayashi y, ohshiro N, Sakai r, et al. transport mechanism and substrate specificity of human organic anion transporter 2 (hoat2 [SLC22A7]). J Pharm Pharmacol. 2005;57(5):573–578.

33. young JD, yao Sy, Sun L, et al. human equilibrative nucleoside trans-porter (eNt) family of nucleoside and nucleobase transporter proteins. Xenobiotica. 2008;38(7–8):995–1021.

34. Poplin e, Wasan h, rolfe L, et al. randomized multicenter, phase II study of Co-101 versus gemcitabine in patients with metastatic pancreatic ductal adenocarcinoma (mPDAC) and a prospective evaluation of the of the asso-ciation between tumor heNt1 expression and clinical outcome with gem-citabine treatment. J Clin Oncol. 2013;13(Suppl):abstract 4007.

Fundingthe study was supported by Cancer research uK (00/006, C245/A82390, SP1984/0204, and SP2590/0101 to JPN). the eSPAC-1 and 3 trials were also supported by the National Cancer Institute of Canada, Canadian Cancer Society; Fonds de recherche de la Société Nationale Française de Gastroentérologie; Fondazione Italiana malattie del Pancreas and Associazione Italiana ricerca Cancro (AIrC regione Veneto n. 6421); health and medical research Council of Australia, Cancer Councils of New South Wales, Queensland, Victoria and South Australia; and the Australasian Gastro-Intestinal trials Group. this work was also partly funded by the Liverpool NIhr Pancreas Biomedical research


unit. JPN is an NIhr Senior Investigator. mm is partially funded by the oxford NIhr Biomedical research Centre.

Notesthe chief investigator, J. P. Neoptolemos, had full access to all the data in the study and takes full responsibility for the integrity of the data and the accu-racy of the data analysis. All of the authors gave final approval of the version to be published. Development of the trial design was by J. P. Neoptolemos, m. W. Büchler, m. J. moore, P. Ghaneh, N. tebbutt, C. Dervenis, A. oláh, F. Laçaine, and r. Carter. these authors led the recruitment with acquisition of data, along with B. Glimelius, J. W. Valle, D. h. Palmer, A. mcDonald, A. Anthoney, A. Scarfe, r. m. Charnley, m. middleton, C. halloran, and J. mayerle. D. h. Palmer was based in Birmingham, uK, during the recruitment phase of the study. All of the aforemen-tioned were responsible for data collection in their own countries and countries of immediate responsibility or sites within countries. F. Campbell was responsi-ble for pathology scrutiny. t. Cox was responsible for detailed statistical analysis and, along with r. Jackson, preparation of reports for the eSPAC group and the monitoring committees. C. L. rawcliffe was the trial coordinator responsible for central administration ensuring ethical standards for collection and verification of data. the results were interpreted by the eSPAC working party (all of the above). W. Greenhalf, J. P. Neoptolemos, D. Palmer, t. Cox, C. L. rawcliffe, P. Ghaneh, e. Costello, and C. halloran prepared the initial draft. r. F. Lamb provided scien-tific input. F. Campbell was responsible for the final editing of the pathology contri-butions. All of the aforementioned were responsible for collating changes into the final paper proposed by participants in the european Study Group for Pancreatic Cancer before final approval. J. mackey and C. Cass provided the heNt1 antibody used; the antibodies were developed in the Cass laboratory and their use as immu-nohistochemistry reagents for cancer was developed by J. mackey.

None of the sponsors or funders influenced the design and conduct of the study, nor the collection, management, analysis, and interpretation of the data, nor the preparation, review, or approval of the manuscript. the authors have no financial disclosures to report. We wish to disclose a com-mercial material transfer Agreement with Clovis and Ventana-roche (Clovis Protocol Number: Co-101-002d, A retrospective Study to evaluate tumor human equilibrative Nucleoside transporter 1 (heNt1) expression and Its relationship to treatment outcome in Patients with Pancreatic Cancer Who have received Gemcitabine) with contract dates from the 28th march 2011

to the 9th october 2011. the work described here is entirely independent of this contract.

technical support was also provided by Deborah J. Latham, North Staffordshire hospital, and rebekah higgins, university of Liverpool. the Independent Data and Safety monitoring Committee was comprised of r.C. G. russell, mD (middlesex hospital, London, uK, Chairman), r. P. Ahern, PhD (Institute for Cancer research, London, uK), and P. Clarke, mD (Clatterbridge Centre for Clinical oncology, Wirral, uK).

Affiliations of authors: Liverpool Cancer Research UK Cancer Trials Unit, Liverpool Cancer Research UK Centre, University of Liverpool, Liverpool, UK (WG, JPN, EG, TFC, PG, EC, CMH, CLR, FC, RJ); the Princess Margaret Hospital, Toronto, Canada (MJM); Manchester Academic Health Sciences Centre, Christie NHS Foundation Trust, School of Cancer and Enabling Sciences, University of Manchester, UK (JWV); Queen Elizabeth Hospital, University Hospital Birmingham NHS Foundation Trust, Birmingham, UK (DHP); Beatson West of Scotland Cancer Centre, Glasgow, UK (ACM); Glasgow Royal Infirmary, Glasgow, UK (RC); Hôpital Tenon, Université, Pierre et Marie Curie, Paris, France (FL); Austin Health, Melbourne, Australia (NCT); Prince of Wales Hospital and Clinical School University of New South Wales, New South Wales, Australia (DG); Nepean Cancer Centre and University of Sydney, Sydney, Australia (JS); Agia Olga Hospital, Athens, Greece (CD); Medical Oncology, Clatterbridge Centre for Oncology, Bebington, Merseyside, UK (DS); Department of Oncology, Akademiska Sjukhuset, Uppsala University, Uppsala, Sweden (BG); University Hospital, North Staffordshire, UK (MD); Freeman Hospital, Newcastle upon Tyne, UK (RMC); Service de Chirurgie Digestive et Viscérale, Hôpital Tenon, Paris, France (FL); Cross Cancer Institute and University of Alberta, Alberta, Canada (JRM, AGS); Churchill Hospital, Oxford University Hospitals NHS Trust, Oxford, UK (MRM); St James’s University Hospital, Leeds, UK (AA); Department of Medicine A, University Medicine Greifswald, Greifswald, Germany (JM); Petz Aladar Hospital, Gyor, Hungary (AO); Departments of Surgery and Pathology and ARC-NET Research Center, University of Verona, Italy (AS, CB); Department of Surgery, University of Heidelberg, Heidelberg, Germany (MWB).

pancreatic cancer hent1 expression and survival from ... · score for their cores was above the...

Documents