Randomized, Double-Blind, Controlled Trial of EarlyEndoscopic Ultrasound–Guided Celiac Plexus Neurolysisto Prevent Pain Progression in Patients With NewlyDiagnosed, Painful, Inoperable Pancreatic CancerJonathan M. Wyse, Marco Carone, Sarto C. Paquin, Mariana Usatii, and Anand V. Sahai

Jonathan M. Wyse, Sarto C. Paquin,Mariana Usatii, and Anand V. Sahai,Centre Hospitalier de l’Universite deMontreal, Hopital Saint Luc, Montreal,Quebec, Canada; and Marco Carone,Johns Hopkins Bloomberg School ofPublic Health, Baltimore, MD.

Submitted August 18, 2010; acceptedJune 14, 2011; published online aheadof print at www.jco.org on August 15,2011.

Presented in part at Digestive DiseaseWeek, May 30-June 4, 2009, Chicago, IL.

Authors’ disclosures of potential con-flicts of interest and author contribu-tions are found at the end of thisarticle.

Clinical Trials repository link available onJCO.org.

Corresponding author: Jonathan Wyse,MDCM, MSc, Centre Hospitalier del’Universite de Montreal, Hopital SaintLuc, 1058 Rue Saint Denis, Montreal,Quebec H2X 3J4, Canada; e-mail:jonwyse@gmail.com.

© 2011 by American Society of ClinicalOncology

0732-183X/11/2926-3541/$20.00

DOI: 10.1200/JCO.2010.32.2750

A B S T R A C T

PurposeCeliac plexus neurolysis (CPN) is currently used as salvage therapy for morphine-resistant pancreatic cancerpain. Endoscopic ultrasound–guided CPN (EUS-CPN) can be performed early, at the time of EUS. Wehypothesized that early EUS-CPN would reduce pain and morphine consumption, increase quality of life(QOL), and prolong survival.

Patients and MethodsPatients were eligible if referred for EUS for suspected pancreatic cancer with related pain. If EUSand EUS-guided fine-needle aspiration cytology confirmed inoperable adenocarcinoma, patientswere randomly assigned to early EUS-CPN or conventional pain management. Pain scores (7-pointLikert scale), morphine equivalent consumption, and QOL scores (Digestive Disease Questionnaire-15)were assessed at 1 and 3 months.

ResultsFive hundred eighty eligible patients were seen between April 2006 and December 2008.Ninety-six patients were randomly assigned (48 patients per study arm). Pain relief was greater inthe EUS-CPN group at 1 month and significantly greater at 3 months (difference in mean percentchange in pain score � �28.9 [95% CI, �67.0 to 2.8], P � .09, and �60.7 [95% CI, �86.6 to�25.5], P � .01, respectively). Morphine consumption was similar in both groups at 1 month(difference in mean change in morphine consumption � �1.0 [95% CI, �47.7 to 49.2], P � .99),but tended toward lower consumption at 3 months in the neurolysis group (difference in meanchange in morphine consumption � �49.5 [95% CI, �127.5 to 7.0], P � .10). There was no effecton QOL or survival.

ConclusionEarly EUS-CPN reduces pain and may moderate morphine consumption in patients with painful,inoperable pancreatic adenocarcinoma. EUS-CPN can be considered in all such patients at the timeof diagnostic and staging EUS.

J Clin Oncol 29:3541-3546. © 2011 by American Society of Clinical Oncology

INTRODUCTION

Pancreatic malignancy is the second highest inci-dent gastrointestinal cancer in North America. In2009, there were an estimated 42,470 new diagno-ses of pancreatic cancer and 35,240 deaths result-ing from the disease.1 Fewer than 20% of patientsare considered operable at diagnosis,2 and theoverall 5-year survival rate is 5%.3 Pain is presentin as many as 70% to 80% of patients at the time ofpancreatic cancer diagnosis,4-6 and can be diffi-cult to control, even with high doses of narcotics,whose adverse effects may further impair qualityof life.7-9

Celiac plexus neurolysis (CPN) is well estab-lished as an effective technique for controlling painand reducing narcotic requirements in patients withpancreatic cancer.6,10 It uses alcohol to permanentlyablate neural tissue of celiac ganglia. However, ran-domized controlled studies of CPN use conven-tional posterior percutaneous or intraoperativemethods,6 which can be invasive and carry a 1% riskof serious adverse effects, including lower extremityparesthesia and paralysis.10

Endoscopic ultrasound (EUS) combineshigh-resolution ultrasound with endoscopy. It iscommonly used to perform EUS-guided fine-needle aspiration cytology of pancreatic masses to

JOURNAL OF CLINICAL ONCOLOGY O R I G I N A L R E P O R T

VOLUME 29 � NUMBER 26 � SEPTEMBER 10 2011

© 2011 by American Society of Clinical Oncology 3541Downloaded from jco.ascopubs.org on September 16, 2012. For personal use only. No other uses without permission.

Copyright © 2011 American Society of Clinical Oncology. All rights reserved.

diagnose pancreatic cancer. During the same endoscopic session, itcan also be used to perform EUS-guided CPN (EUS-CPN) via atrans-gastric approach. Under EUS guidance, celiac ganglia can belocated, and neurolytic substances such as absolute alcohol can beinjected precisely and safely into and around the celiac ganglia.EUS-CPN has displayed both encouraging results and an excellentsafety profile in several studies.11-16 In a meta-analysis of EUS-CPNstudies with varying length of follow-up, pain reduction was ob-served in approximately 80% of patients.13

CPN is currently used late in disease evolution, as salvage therapyfor morphine-resistant pancreatic cancer pain. Because EUS is used atthe beginning of the diagnostic algorithm for pancreatic cancer, itprovides an opportunity to perform EUS-CPN early (at the time ofdiagnosis). We hypothesized that, as compared with conventionalpain management with narcotics alone, early EUS-CPN would pre-vent progression of pain and narcotic use in patients with painful,inoperable pancreatic cancer.

PATIENTS AND METHODS

Trial Design

This randomized, double-blind, controlled clinical trial was conductedfrom April 2006 to December 2008 at the Centre Hospitalier de l’Universite deMontreal in Montreal, Quebec, Canada. The study protocol was approved bythe institutional review board. It was registered with ClinicalTrials.gov. Thetrial was registered after the trial was started, because universal registration ofclinical trials had only just become the norm. All patients provided written,informed consent before EUS and possible sedation.

All patients 18 years or older referred for EUS for diagnosis and staging ofsuspected pancreatic cancer were approached by a research assistant who

explained the study protocol and determined whether the patient met eligibil-ity criteria. To be eligible, patients were required to have suspected pancreaticcancer and any new-onset pain considered to be cancer-related (centrallylocated, constant, with no other obvious cause). Exclusion criteria were anallergy to bupivacaine, possible future surgical management of the tumor,expected survival less than 3 months (suspected or proven carcinomatosisand/or liver metastases), and inability or unwillingness to provide in-formed consent.

All patients were administered a questionnaire to assess demographics,gastrointestinal symptoms, narcotic use, abdominal pain, and quality of life.Pain levels were quantified using a 7-point Likert scale. Patients were asked tograde their level of pain from 0 to 6 according to Appendix Table A1 (onlineonly). A visual analog scale of pain severity was not feasible because our patientpopulation was unable to reliably present for frequent follow-up to our clinic,and phone interviews were easily obtained. Furthermore, Likert pain scales areoften used and have been validated.17-21 Quality of life was quantified using thevalidated Digestive Disease Questionnaire-15 (DDQ-15) quality-of-life in-strument (Appendix Table A2, online only).22 Diagnostic and staging EUS wasthen performed as an outpatient procedure under conscious sedation, usingtherapeutic, linear array echoendoscope (Pentax, 3830UT, Pentax America,Montvale, NJ), by one of two experienced endosonographers (� 5,000 proce-dures each). If a cytologic diagnosis of adenocarcinoma was obtained (byon-site cytopathology of EUS-guided fine-needle aspiration) and the lesionwas deemed inoperable, then the patient was randomly assigned. Inoperabilitywas based on other available imaging and/or the EUS findings and was con-firmed by per-procedure surgical consultation. EUS findings of inoperabilitywere any of (1) involvement of more than 50% of the circumference of thesuperior mesenteric vein, portal vein, or spleno-portal confluence; (2) anyinvolvement of superior mesenteric artery, celiac axis, or hepatic artery; or (3)cytologically proven nonregional lymphadenopathy.

Patients were randomly assigned by sealed envelope during the proce-dure to EUS-CPN or no EUS-CPN. A computer-generated number sequencewas used to establish random allocation. In patients randomly assigned to

Assessed for eligibility (N = 580)

Excluded

Did not meet inclusion criteriaRefused to participateOther reasons

(n = 482)

(n = 434)(n = 41)(n = 7)E

nro

llmen

tA

lloca

tio

nFo

llow

-up

An

alys

is

Randomly allocated (n = 98)

Allocated to intervention

Received interventionDid not receive intervention

Lost to follow-up

Discontinued intervention

Analyzed

Excluded from analysis

Allocated to intervention

Received interventionDid not receive intervention

(n = 49)(n = 0)

Lost to follow-up

Discontinued intervention

(n = 4)

(n = 0)

Analyzed

Excluded from analysis

(n = 49)

(n = 0)

(n = 98)

(n = 49)(n = 0)

(n = 2)

(n = 0)

(n = 49)

(n = 0)

Control Neurolysis

(n = 98)

Fig 1. CONSORT diagram.

Wyse et al

3542 © 2011 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on September 16, 2012. For personal use only. No other uses without permission.Copyright © 2011 American Society of Clinical Oncology. All rights reserved.

EUS-CPN, the technique was performed immediately using a 19-gauge needle(Echotip 19, Cook Medical, Winston-Salem, NC) with bilateral injectionaround the celiac axis with a total of 10 mL of 0.5% bupivacaine and 20 mL ofabsolute alcohol as previously described.14 Afterward, patients were returnedto their referring physicians, who made all further pain management decisions.Beyond 1 month after randomization, patients were permitted to undergoopen-label EUS-CPN at the discretion of their referring physician.

Postprocedure questionnaires were administered by telephone by thesame research assistant who initially evaluated the patient. The patients, refer-ring physicians, and the research assistant were blinded to treatment arm. Theresearch assistant recorded all outcomes and was not present in the roomduring the EUS procedure.

Primary and Secondary Outcomes

All outcomes were assessed at 1 and 3 months after randomization. Eachpatient served as his/her own control. The primary study outcome was painrelief for each patient, at 1 and 3 months, quantified as percent change inabdominal pain score at 1 month and 3 months, as compared with the pa-tient’s baseline pain score.

Secondary outcomes were absolute change in Likert pain score, absolutechange in morphine use (expressed as the change in morphine equivalent[MEQ] consumption compared with patient’s baseline MEQ consumption),change in quality-of-life (DDQ-15) scores, and overall survival. Furtherplanned secondary analysis was performed based on whether patients receivedpalliative chemotherapy and/or radiation therapy (chemo-XRT).

Complications

Complications were defined as any unplanned event considered relatedto EUS-CPN that required prolongation of hospitalization immediately afterthe procedure (early), or after discharge (late).

Sample Size Calculation

Preliminary calculations indicated that a sample of 84 patients (equallydivided into control and treatment groups) would suffice to achieve 80%statistical power for detecting a 30% difference in mean relative reduction ofpain scores between treatment arms after 3 months, at a two-sided significancelevel of .05. For these calculations, patients undergoing diagnostic EUS alonewere assumed to themselves experience a 30% decrease in pain scores (com-pared with their baseline pain scores) due to a placebo effect. These relativechanges were assumed to have a common standard deviation between 30%and 35%, and the aggregate attrition rate at the 3-month mark (including bothmortality and loss to follow-up) was assumed to be approximately 50%.Relative changes in pain scores for the conventional management and exper-imental arms were assumed independent, with nonskewed distributions.

Statistical Analysis

Data were analyzed on an “intent to treat” basis. The study ended whenall patients were either deceased or had reached 3 months of follow-up. Atevery time point, analysis of changes compared with baseline included dataavailable for all patients still under study at this time point. Mean differenceswere expressed as coefficients in appropriate linear models and estimatedusing generalized estimating equations23,24; this allowed the use of data fromdiffering time points in the estimation procedure. CIs for these mean differ-ences and hypothesis tests comparing outcomes between treatment groupswere constructed using Efron’s bias-corrected accelerated nonparametricbootstrap approach (Boot: Bootstrap R [S-Plus] Functions, R package version1.2).25,26 Although computationally more demanding, this method generallyprovides more exact inference, particularly when a normal approximation tothe finite-sample distribution of the test statistic is suspected not to be ade-quate.25 Analyses of change in outcomes at 1 and 3 months relative to baselinewere performed marginally. Kaplan-Meier estimates of the survival curve wereobtained for each treatment group and compared using the log-rank test.27,28

Cox proportional hazards models were used to study survival while adjustingfor potential confounding factors.29 All analyses were performed using the Rstatistical package (version 2.8.1).30 Data unavailable for analysis were as-sumed to be missing completely at random; similarly, patients lost tofollow-up were assumed to be similar to other members of their respectivetreatment group. All statistical tests of hypothesis were two-sided and con-

ducted at significance level .05. Ninety-five percent CIs are provided in brack-ets after all point estimates.

RESULTS

Between April 2006 and December 2008, 580 patients were referredfor EUS for suspected pancreatic cancer (Fig 1). Four hundred eighty-two patients did not meet study entry criteria (did not have biopsy-proven cancer, had cancer but either was possibly operable and/or hadno pain) or refused to participate. Ninety-eight patients were ran-domly assigned, with 49 patients randomly assigned to each arm.Patients in both groups were comparable for all cogent variables (Ta-ble 1). Effective sample size for variables at different time points isdisplayed in Appendix Table A3 (online only). All patients had locore-gional disease. Patients who received EUS-CPN showed no evidenceof early or late complications. No patient from the CPN group re-ceived open-label CPN. Two patients from the control group receivedopen-label CPN at 43 and 52 days. Two patients (one from each arm)were later revealed ineligible as a result of chemotherapy and/or radi-ation therapy received before random assignment. Five patients werelost to follow-up before 1 month (four in the neurolysis group [meanbaseline pain score � 5.5] and one in the control group [baseline painscore � 6]). One patient from the control group was lost to follow-upbetween 1 and 3 months (baseline pain score � 4). All patients wereincluded in the intention-to-treat analysis. Of 82 patients surviving to1 month, pain scores and quality-of-life scores were not obtained intwo and three patients, respectively. Of 60 patients surviving to 3

Table 1. Baseline Characteristics of All Randomly Assigned Patients

Characteristic Control Treatment

No. of patients 49 49Age, years

Mean 66.5 66.6SD 10.0 9.32

Male sexNo. 21 26% 42.9 53.1

Pain history, weeksMean 9.6 8.6SD 8.4 5.4

Narcotic consumption, morphine-equivalentunits

Mean 36.8 41.9SD 61.7 74.1

Abdominal pain intensity, Likert scoreMean 4.1 5.2SD 1.6 1.6

Quality of life, Digestive DiseaseQuestionnaire score

Mean 3.1 3.1SD 0.7 0.7

ConstipationNo. 18 of 44 25 of 46% 40.9 54.3

NauseaNo. 23 of 47 21 of 47% 48.9 44.7

Abbreviation: SD, standard deviation.

Endoscopic Ultrasound Celiac Neurolysis for Pancreatic Cancer

www.jco.org © 2011 by American Society of Clinical Oncology 3543Downloaded from jco.ascopubs.org on September 16, 2012. For personal use only. No other uses without permission.

Copyright © 2011 American Society of Clinical Oncology. All rights reserved.

months, pain scores, MEQ consumption, and quality-of-life scoreswere not obtained in two, one, and six patients, respectively.

Primary Outcome: Percent Change in Pain Scores

Relative to Baseline

In the control group, pain scores tended to increase comparedwith baseline at both 1 month (mean percent change in pain score ��11 [95% CI, �13 to �49]) and 3 months (mean percent change inpain score � �12 [95% CI, �19 to �36]). The absolute mean changein pain scores in the control group were�0.6 (95% CI,�1.1 to�0.09)at T1 and �0.3 (95% CI, �0.9 to �0.2) at T3. In contrast, in theEUS-CPN group, pain scores tended to decrease compared with base-line at 1 month (mean percent change in pain score � �18 [95% CI,�37 to 0]) and decreased significantly at 3 months (mean percentchange in pain score � �49 [95% CI, �61 to �38]). The absolutemean change in pain scores in the EUS-CPN group was significant atboth time points: �1.6 (95% CI, �2.1 to �1.0) at T1 and �2.6 (95%CI, �3.2 to �2.0) at T3. When comparing both groups, pain reliefwith EUS-CPN was greater at 1 month (difference in mean percentchange in pain score � �28.9 [95% CI, �67.0 to 2.8], P � .09) andsignificantly greater at 3 months after intervention (difference inmean percent change in pain score � �60.7 [95% CI, �86.6 to�25.5], P � .01; Fig 2A). The difference in absolute mean changewas significantly lower in the CPN-EUS group at both T1 (�1.0[95% CI, �1.7 to �0.1], P � .01) and T3 (�2.2 [95% CI, �3.1 to�1.4], P � .001; Fig 2B).

Secondary Outcomes

Morphine use. In the control group, morphine use increasedcompared with baseline at both 1 month (mean absolute change inMEQ consumption��54 [95% CI,�20 to�96]) and particularly at3 months (mean absolute change in MEQ consumption � �100[95% CI, �49 to �180]). In the EUS-CPN group, morphine use alsoincreased at 1 month (mean change in MEQ consumption � �53[95% CI, �28 to �89]), but seemed to plateau at 3 months (meanchange in MEQ consumption � �50 [95% CI, �28 to �79]). How-ever, the differences between the control and EUS-CPN groups werenot significant at 1 month (difference in mean change in morphineconsumption � �1.0 [95% CI, �47.7 to 49.2], P � .99) or 3 months(difference mean change in morphine consumption � �49.5 [95%CI, �127.5 to 7.0], P � .10; Fig 3).

Quality of life. In the control group, quality-of-life scores tendedto increase compared with baseline at 1 month (mean percentchange � � 6 [95% CI, 0 to �11]) and 3 months (mean percentchange � �18 [95% CI, �12 to �26]). Quality of life also seemed toincrease in the EUS-CPN group at 1 month (mean percent change ��14 [95% CI, �7 to �19]) and 3 months (mean percent change ��19 [95% CI, �10 to �27]). The differences between the two groupswere not significant at 1 month or 3 months.

Survival. There was no difference between groups in overallsurvival (P � .90).

Planned Secondary Analyses

Effects of chemo-XRT. Because chemo-XRT is known to im-prove pain and quality of life in advanced pancreatic cancer, a plannedsecondary analysis was stratified for this variable. In patients who didnot undergo chemo-XRT, EUS-CPN was associated with significantlybetter pain relief at both 1 month (difference in mean percentchange � �59.6 [95% CI, �95.4 to �27.6], P � .001) and 3 months(difference in mean percent change � �85.8 [95% CI, �127.6 to�51.3], P � .001). Similarly, the difference in absolute mean painscores was �1.7 (95% CI, �2.6 to �0.7; P � .007) at T1 and �2.2

After 1 month After 3 months

P = .085 P = .011

A

Aver

age

Perc

ent C

hang

e in

Abd

omin

al

Like

rt Sc

ore

Com

pare

d W

ith B

asel

ine

0

-100

100

200

300

400

500

600

After 1 month After 3 months

P = .011 P < .001

B

Aver

age

Abso

lute

Cha

nge

in A

bdom

inal

Li

kert

Scor

e Co

mpa

red

With

Bas

elin

e

-4

-6

-2

0

2

4

6ControlNeurolysis

ControlNeurolysis

Fig 2. (A) Abdominal pain scores, per-cent change. (B) Abdominal pain scores,absolute change.

After 1 month After 3 months

P = .987 P = .101

Aver

age

Chan

ge in

Pai

nkill

er C

onsu

mpt

ion

(in m

orph

ine−

equi

vale

nt u

nits

)Co

mpa

red

With

Bas

elin

e

0

-100

-200

100

200

300

400

500

600 ControlNeurolysis

Fig 3. Morphine equivalent consumption.

Wyse et al

3544 © 2011 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on September 16, 2012. For personal use only. No other uses without permission.Copyright © 2011 American Society of Clinical Oncology. All rights reserved.

(95% CI, �3.7 to �0.3; P � .03) at T3. In patients who underwentchemo-XRT, EUS-CPN was not significantly different from control atT1 (difference in mean percent change � 31.0 [95% CI, �34.3 to106.2]; P � .38), but was significantly better at 3 months (difference inmean percent change � �45.6 [95% CI, �72.6 to �23.3]; P � .001;Fig 4A). Similarly, the difference in absolute mean pain scores was 0.5(95% CI, �1.1 to 2.2; P � .48) at T1 and �1.9 (95% CI, �3.0 to �0.7;P � .009) at T3 (not shown).

MEQ consumption. In those who did not undergo chemo-XRT,differences in MEQ consumption were not significant at 1 month(difference in mean change in morphine consumption � �2.4 [95%CI, �58.4 to 60.8]; P � .97) but significantly favored neurolysis at 3months (difference in mean change in morphine consumption ��144.5 [95% CI, �290.9 to �30.0]; P � .05). In patients undergoingchemo-XRT, the difference between groups were not significant at 1month (difference in mean change in morphine consumption � 11.4[95% CI, �23.7 to 39.4]; P � .49) or 3 months (difference meanchange in morphine consumption � 26.1 [95% CI, �12.2 to 56.5],P � .16; Fig 4B).

DISCUSSION

Given the dismal prognosis of inoperable pancreatic cancer andpotential for intractable, narcotic-dependent pain, any low-risktherapy that may mitigate these symptoms is likely to have animportant role in patient management. Uncontrolled trials suggestthat EUS-CPN provides relief as salvage therapy for intractable,

narcotic-resistant cancer-related pain in patients with inoperablepancreatic cancer. We hypothesized that, compared with conven-tional management, early EUS-CPN for cancer-related pain wouldprevent progression of pain and narcotic use, improve quality oflife, and perhaps prolong survival.

Our results show that, compared with conventional manage-ment with narcotics alone, early EUS-CPN reduces pain to a greaterextent, without a concomitant increase in narcotic consumption. Ef-fects on pain and narcotic consumption were seen primarily later indisease evolution (at 3 months after randomization). This is not un-expected, because cancer-related pain worsens with time, and benefitsof early neurolysis become evident only after the cycle of pain andnarcotic consumption has time to progress. Although not our primaryoutcome, absolute change in Likert pain score significantly favoredEUS-CPN at both 1 month and 3 months. When considering themeta-analysis by Yan and Myers6 of computed tomography–guidedand intraoperative CPN, at 1 and 2 months the percent change inpain scores were 10% and 12%, respectively. Our study demon-strated 1- and 3-month percent change in pain scores of 18% and49%, respectively. This greater reduction in pain is more likelyclinically significant. Stratification for chemo-XRT showed thatearly EUS-CPN was particularly effective in providing pain reliefand preventing progressive increased narcotic consumption inpatients who did not undergo chemo-XRT.

Both control and CPN groups demonstrated increased mor-phine consumption relative to baseline at 1 month. However,although morphine consumption stabilized in the CPN group, it

After 1 month After 3 months

After 1 month After 3 months

P < .001 P < .001

A

B

Aver

age

Chan

ge in

Abd

omin

al L

iker

t Sc

ore

Com

pare

d W

ith B

asel

ine

(%)

0

-100

100

200

300

400

500

600

Aver

age

Chan

ge in

Abd

omin

al L

iker

t Sc

ore

Com

pare

d W

ith B

asel

ine

(%)

ControlNeurolysis

After 1 month After 3 months

P = .384 P < .001

0

-100

100

200

300

400

500

600ControlNeurolysis

P = .969 P = .005

Aver

age

Abso

lute

Cha

nge

in P

aink

iller

Co

nsum

ptio

n (in

mor

phin

e−eq

uiva

lent

uni

ts)

Com

pare

d W

ith B

asel

ine

0

-100

-200

100

200

300

400

500

600

After 1 month After 3 months

P = .490 P = .159

Aver

age

Abso

lute

Cha

nge

in P

aink

iller

Co

nsum

ptio

n (in

mor

phin

e−eq

uiva

lent

uni

ts)

Com

pare

d W

ith B

asel

ine

0

-100

-200

100

200

300

400

500

600

Fig 4. (A) Percent change in pain scoresstratified by chemotherapy plus radiation(chemo-XRT) status. (B) Change in mor-phine equivalent consumption stratifiedby chemo-XRT status. Patients receivingno chemo-XRT, left column; patients re-ceiving chemo-XRT, right column.

Endoscopic Ultrasound Celiac Neurolysis for Pancreatic Cancer

www.jco.org © 2011 by American Society of Clinical Oncology 3545Downloaded from jco.ascopubs.org on September 16, 2012. For personal use only. No other uses without permission.

Copyright © 2011 American Society of Clinical Oncology. All rights reserved.

continued to increase in the control group, resulting in a between-group difference that trended in favor of CPN (difference meanchange in morphine consumption � �49.5 [95% CI, �127.5 to7.0]). We believe lack of significance reflects being slightly under-powered for this secondary outcome.

Despite better pain control, early EUS-CPN did not produce ademonstrable improvement in quality of life, or in survival. Lack ofimpact on quality of life could be explained by the instrument used,DDQ-15, which does not include a pain domain. We specifically chosea quality of life instrument without a pain domain, because pain reliefwas already the primary study outcome. The study was also not pow-ered to look for effects on quality of life.

Alternative statistical methods, including, for example, models forjoint survival and longitudinal data, and propensity score approachescould have been used. However, a scrutiny of the substantive basis of ourstudy, both theoretical and empirical, did not seem to suggest clear bene-fitsofmorecomplexstatisticalmethodology,whereastheirusecouldhavepossibly compromised ease of interpretation of results.

Early EUS-CPN is safe (no early or late complications) and addsapproximately 2 to 3 minutes to the diagnostic and staging EUSprocedure. Still, one could argue that early EUS-CPN should be re-served only for patients who are not candidates for chemo-XRT.However, at the time of diagnostic and staging EUS, it is usuallyunclear whether patients are candidates for chemo-XRT, would ac-cept chemo-XRT, or can complete chemo-XRT, once initiated.

Therefore, we conclude that in patients with painful, inoperablepancreatic adenocarcinoma, early EUS-CPN provides better pain reliefand may prevent progressive increases in morphine consumption com-paredwithconventionalmanagement,particularlyinpatientswhodonotreceive chemo-XRT. Because EUS-CPN is safe and it is often unclearwhich patients will receive and/or tolerate chemo-XRT, we recommendearly EUS-CPN be considered during diagnostic and staging EUS in allpatients with predicted survival of several months in whom confirmationof painful, locoregional, and inoperable pancreatic cancer is obtained.

AUTHORS’ DISCLOSURES OF POTENTIAL CONFLICTSOF INTEREST

The author(s) indicated no potential conflicts of interest.

AUTHOR CONTRIBUTIONS

Conception and design: Anand V. SahaiAdministrative support: Mariana Usatii, Anand V. SahaiProvision of study materials or patients: Sarto C. Paquin,Anand V. SahaiCollection and assembly of data: Jonathan M. Wyse, Sarto C. Paquin,Mariana Usatii, Anand V. SahaiData analysis and interpretation: Jonathan M. Wyse, Marco Carone,Sarto C. Paquin, Anand V. SahaiManuscript writing: All authorsFinal approval of manuscript: All authors

REFERENCES

1. American Cancer Society: Cancer Facts & Figures2009. Atlanta, GA, American Cancer Society, 2009

2. Bilimoria KY, Bentrem DJ, Ko CY, et al: Vali-dation of the 6th edition AJCC Pancreatic CancerStaging System: Report from the National CancerDatabase. Cancer 110:738-744, 2007

3. Jemal A, Siegel R, Ward E, et al: Cancerstatistics, 2008. CA Cancer J Clin 58:71-96, 2008

4. de Oliveira R, dos Reis MP, Prado WA: Theeffects of early or late neurolytic sympathetic plexusblock on the management of abdominal or pelviccancer pain. Pain 110:400-408, 2004

5. Wong GY, Schroeder DR, Carns PE, et al:Effect of neurolytic celiac plexus block on pain relief,quality of life, and survival in patients with unresect-able pancreatic cancer: A randomized controlledtrial. JAMA 291:1092-1099, 2004

6. Yan BM, Myers RP: Neurolytic celiac plexusblock for pain control in unresectable pancreaticcancer. Am J Gastroenterol 102:430-438, 2007

7. Collins D, Penman I, Mishra G, et al: EUS-guided celiac block and neurolysis. Endoscopy 38:935-939, 2006

8. Michaels AJ, Draganov PV: Endoscopic ultra-sonography guided celiac plexus neurolysis andceliac plexus block in the management of pain dueto pancreatic cancer and chronic pancreatitis. WorldJ Gastroenterol 13:3575-3580, 2007

9. Schmulewitz N, Hawes R: EUS-guided celiacplexus neurolysis–technique and indication. Endos-copy 35:S49-53, 2003

10. Eisenberg E, Carr DB, Chalmers TC: Neuro-lytic celiac plexus block for treatment of cancer pain:A meta-analysis. Anesth Analg 80:290-295, 1995

11. Gunaratnam NT, Sarma AV, Norton ID, et al: Aprospective study of EUS-guided celiac plexus neu-rolysis for pancreatic cancer pain. Gastrointest En-dosc 54:316-324, 2001

12. Levy MJ, Topazian MD, Wiersema MJ, et al:Initial evaluation of the efficacy and safety of endo-scopic ultrasound-guided direct Ganglia neurolysisand block. Am J Gastroenterol 103:98-103, 2008

13. Puli SR, Reddy JB, Bechtold ML, et al: EUS-guided celiac plexus neurolysis for pain due tochronic pancreatitis or pancreatic cancer pain: Ameta-analysis and systematic review. Dig Dis Sci54:2330-2337, 2009

14. Sahai AV, Lemelin V, Lam E, et al: Central vs.bilateral endoscopic ultrasound-guided celiac plexusblock or neurolysis: A comparative study of short-termeffectiveness. Am J Gastroenterol 104:326-329, 2009

15. Tran QN, Urayama S, Meyers FJ: Endoscopicultrasound-guided celiac plexus neurolysis for pancreaticcancer pain: A single-institution experience and review ofthe literature. J Support Oncol 4:460-462, 464, 2006

16. Wiersema MJ, Wiersema LM: Endosonography-guided celiac plexus neurolysis. Gastrointest Endosc44:656-662, 1996

17. Impellizzeri FM, Maffiuletti NA: Convergentevidence for construct validity of a 7-point likertscale of lower limb muscle soreness. Clin J SportMed 17:494-496, 2007

18. Johansson M, Finizia C, Innocenti AD, et al: Is aline the same as a box? Speech assessment by VAS isnot superior to Likert scales in laryngeal cancer patients.Med Sci Monit 13:CR481-CR487, 2007

19. Vickers AJ: Comparison of an ordinal and acontinuous outcome measure of muscle soreness.Int J Technol Assess Health Care 15:709-716, 1999

20. Singh RH, Ghanem KG, Burke AE, et al:Predictors and perception of pain in women under-going first trimester surgical abortion. Contraception78:155-161, 2008

21. Stacey BR, Glanzman RL: Use of gabapentin forpostherpetic neuralgia: Results of two randomized,placebo-controlled studies. Clin Ther 25:2597-2608, 2003

22. Hebert RL, Palesch YY, Mauldin PD, et al: DDQ-15Health-Related Quality of Life Instrument for Patientswith Digestive Disorders. Health Serv Outcomes ResMethodol 2:137-156, 2001

23. Liang KY, Zeger SL: Longitudinal data analysisusing generalized linear models. Biometrika 73:13-22,1986

24. Højsgaard S, Halekoh U, Yan J: The R Packagegeepack for Generalized Estimating Equations. JStat Softw 15:1-11, 2005

25. Efron B: Better bootstrap confidence inter-vals. J Am Stat Assoc 397:171-185, 1987

26. Reference deleted27. Kaplan EL, Meier P: Nonparametric estima-

tion from incomplete observations. J Am Stat Assoc53:457-481, 1958

28. Mantel N: Evaluation of survival data and twonew rank order statistics arising in its consideration.Cancer Chemother Rep 50:163-170, 1966

29. Cox DR: Regression models and life tables. JR Stat Soc B 34:187-220, 1972

30. R Development Core Team: R: A language andenvironment for statistical computing. Vienna, Austria, RFoundation for Statistical Computing, 2008

■ ■ ■

Wyse et al

3546 © 2011 by American Society of Clinical Oncology JOURNAL OF CLINICAL ONCOLOGY

Downloaded from jco.ascopubs.org on September 16, 2012. For personal use only. No other uses without permission.Copyright © 2011 American Society of Clinical Oncology. All rights reserved.