Date post: | 05-Sep-2016 |
Category: |
Documents |
Upload: | roger-chou |
View: | 212 times |
Download: | 0 times |
1026 Journal of Pain and Symptom Management Vol. 26 No. 5 November 2003
Review Article
Comparative Efficacy and Safetyof Long-Acting Oral Opioids for ChronicNon-Cancer Pain: A Systematic ReviewRoger Chou, MD, Elizabeth Clark, MD, MPH, and Mark Helfand, MD, MPHDepartments of Medicine (R.C., M.H.) and Family Medicine (E.C.), Oregon Health & ScienceUniversity; Oregon Evidence-Based Practice Center (R.C., M.H., E.C.); and Portland VeteransAffairs Medical Center (M.H.), Portland, Oregon, USA
AbstractOpioids have been endorsed as appropriate treatment for refractory chronic non-cancer painwhen used according to published guidelines. They are widely used for this indication.However, there appear to be gaps in our understanding of the efficacy and safety ofindividual long-acting opioids compared to each other or as a class compared to short-acting opioids. This systematic review summarizes and assesses the evidence for thecomparative efficacy and safety of long-acting opioids in the management of chronic non-cancer pain. Randomized trials (for comparative efficacy and adverse events) andobservational studies (for adverse events only) that included non-parenteral long-actingopioids were sought using electronic databases, handsearching reference lists, and solicitingpharmaceutical company submissions. Searches were performed through October 2002. Thevalidity of each included study was assessed using a data abstraction form and predefinedcriteria. An overall grade was allocated for the body of evidence for each key question. Atotal of 16 randomized trials (comparative efficacy and adverse events), enrolling 1427patients, and 8 observational studies (adverse events) of 1190 patients were included inthis review. No randomized trial was rated good quality; observational studies weregenerally of poorer quality than the trials. There was insufficient evidence to prove thatdifferent long-acting opioids are associated with different efficacy or safety profiles. Therewas also insufficient evidence to determine whether long-acting opioids as a class are moreeffective or safer than short-acting opioids. A subgroup of three studies on long-actingversus short-acting oxycodone was more homogeneous and provided fair evidence that theseformulations are equally effective for pain control. J Pain Symptom Manage2003;26:1026–1048. � 2003 U.S. Cancer Pain Relief Committee. Published by ElsevierInc. All rights reserved.
Key WordsAnalgesics, opioid, pain, meta-analysis, fentanyl, morphine, oxycodone, codeine
Address reprint requests to: Roger Chou, MD, 9721 SWMorrison St., Portland, OR 97225, USA.Accepted for publication: March 18, 2003.
� 2003 U.S. Cancer Pain Relief CommitteePublished by Elsevier Inc. All rights reserved.
Note: Evidence tables and appendices are avail-able on the website http://www.oregonrx.org/OrgrxPDF/Opioid%20Review.htm (last updated April2002) or from the authors (updated October 2002).
0885-3924/03/$–see front matterdoi:10.1016/j.jpainsymman.2003.03.003
Vol. 26 No. 5 November 2003 1027Long-Acting Oral Opioids for Chronic Non-Cancer Pain
IntroductionChronic pain, typically defined as pain of at
least 6 months’ duration, is a common causeof major disability. It is estimated that 1 in 5adult Americans, or 30 million people, experi-ence chronic pain.1 Chronic non-cancer painafflicts a significant subset of chronic pain pa-tients, causing personal suffering, reduced pro-ductivity, and substantial health care costs.2
Opioids have been endorsed by the AmericanAcademy of Pain Medicine and the AmericanPain Society3 as appropriate treatment for re-fractory chronic non-cancer pain in the generalpopulation as well as in older patients,4 whenused judiciously and according to guidelinessimilar to those used for cancer patients.
Opioids are a class of medications that acton common receptors and are natural deriva-tives of morphine.5 They are the most potentmedications available for treatment of mosttypes of severe pain. They are also associatedwith a variety of adverse events, including abuseand addiction. Opioids are available in bothshort- and long-acting preparations, and the useof long-acting opioids for patients with chronicnon-cancer pain has become common. Becausechronic pain may not resolve over time, use ofopioid analgesics for these conditions can belong-term. Despite the widespread use of long-acting opioids, there are few data regarding thecomparative efficacy and adverse event profilesassociated with specific long-acting opioids inpatients who have chronic non-cancer pain.6
In 2001, the Oregon Legislature passedSenate Bill 819, which mandated the develop-ment of a Practitioner-Managed PrescriptionDrug Plan (PMPDP) for the Oregon HealthPlan (OHP). The Oregon Health Plan refersto a collective series of laws enacted from 1989through 1995 that sought to expand Medicaidcoverage to low-income Oregonians by creatingstate-run insurance pools, enacting insurancereforms, using a federal waiver that allowed forMedicaid expansion, and excluding certain di-agnoses and treatments from coverage. As partof this process for developing a PMPDP forthe Oregon Health Plan, the Oregon HealthResources Commission (OHRC) required thatan evidence-based review of the state’s most ex-pensive drug classes be performed. The OHRCrequested a review of the long-acting opioiddrug class specifically in persons with chronic
non-cancer pain. The OHRC requested infor-mation about whether there is evidence thatone or more long-acting opioid is superior toothers in terms of efficacy and safety, andalso whether long-acting opioids as a class aremore efficacious or safer than short-acting opi-oids in the treatment of chronic non-cancerpain.
Scope and Key QuestionsThe scope of the review and key questions
were developed and refined with input from anOHRC subcommittee comprised of statewideexperts (pharmacists, primary care clinicians,pain care specialists, and representatives of thepublic). In consultation with the subcommit-tee, we selected the following key questions toguide the review:
1. What is the comparative efficacy of differ-ent long-acting opioids in reducing painand improving functional outcomes inadult patients being treated for chronicnon-cancer pain?
A. In head-to-head comparisons, hasone or more long-acting opioidbeen shown to be superior to otherlong-acting opioids in reducingpain and improving functional out-comes when used for treatment ofadults with chronic non-cancerpain?
B. In trials comparing long-acting opi-oids to other types of drugs or toplacebo, is there a pattern to suggestthat one long-acting opioid is moreeffective than another?
C. Have long-acting opioids beenshown to be superior to short-actingopioids in reducing pain and im-proving functional outcomes whenused for treatment of adults withchronic non-cancer pain?
2. What are the comparative incidence andnature of adverse effects (including addic-tion and abuse) of long-acting opioidmedications in adult patients beingtreated for chronic non-cancer pain?
A. In head-to-head comparisons, hasone or more long-acting opioidbeen shown to be associated with
1028 Vol. 26 No. 5 November 2003Chou et al.
fewer adverse events compared toother long-acting opioids whenused for treatment of adults withchronic non-cancer pain?
B. In trials comparing long-acting opi-oids to other types of drugs or toplacebo, is there a pattern to suggestthat one long-acting opioid is associ-ated with fewer adverse eventsthan another?
C. Have long-acting opioids beenshown to have fewer adverse eventsthan short-acting opioids when usedfor treatment of adults with chronicnon-cancer pain?
3. Are there subpopulations of patients (spe-cifically by race, age, sex, or type of pain)with chronic non-cancer pain for whichone long-acting opioid is more effectiveor associated with fewer adverse effects?
Several aspects of the key questions deservecomment:
Population. The population included in thisreview is adult (greater than 18 years old) pa-tients with chronic non-cancer pain. We de-fined chronic non-cancer pain as continuousor recurring pain of at least 6 months’ duration.Senate Bill 819 specifically excludes cancer pa-tients and patients with HIV from the PMPDPprocess, and they were not part of this review.
Drugs. We included oral or transdermal long-acting opioids. “Long-acting” was defined asopioids administered twice a day or less fre-quently. Long-acting opioids that we identifiedwere transdermal fentanyl and oral oxycodone,morphine, methadone, levorphanol, codeine,and dihydrocodeine. “Short-acting” wasdefined as opioids administered more fre-quently than three times a day. Although “sus-tained-release” and “immediate-release” areother terms used to describe the onset and du-ration of action of certain opioid preparations,for this review we classified opioids on thebasis of dosing frequency.
Outcomes. The main efficacy measures werepain intensity, pain relief, and function. There isno single accepted standard regarding how tomeasure these outcomes.
Most studies measure pain intensity usingeither visual analogue or categorical pain scales.Visual analogue scales (VAS) consist of a lineon a piece of paper labeled 0 at one end, indi-cating no pain, and a maximum number (com-monly 100) at the other, indicating excruciatingor most severe pain. Patients designate theircurrent pain level on the line. An advantage ofVAS is that they provide a continuous rangeof values for relative severity. A disadvantage isthat the meaning of a pain score for any individ-ual patient remains arbitrary. Categorical painscales, on the other hand, consist of severalpain category options from which a patientmust choose (e.g., no pain, mild, moderate, orsevere). A disadvantage of categorical scales isthat patients must chose between categories thatmay not accurately describe their pain. Thebest approach may be to utilize both methods.7
Pain control (improvement in pain) and painrelief (resolution of pain) are also measuredusing visual analogue and categorical scales.
Studies usually evaluate function using theMedical Outcomes Study Short Form-36 (SF-36), Short Form-12 (SF-12), or another multi-question assessment. These questionnairesmeasure how well an individual functions physi-cally, socially, cognitively, and psychologically.Another approach to measuring function is tofocus on how well the medication helps prob-lems in daily living commonly faced by patientswith chronic pain, such as getting enough sleepor staying focused on the job. Some studies alsoreport effects on mood and the preference forone medication over another.
The subcommittee selected the following ad-verse events for our review: abuse, addiction,respiratory depression, nausea, vomiting, con-stipation, dizziness, somnolence, and confu-sion. These were the adverse events felt bythe subcommittee to be the most common ortroubling adverse events in clinical practice. Werecorded rates of these adverse events as wellas rates of discontinuation due to a particularadverse effect. In some studies, only “serious”adverse events or adverse events “thought re-lated to treatment medication” are reported.Many studies do not define these terms.
The subcommittee specifically requested thatwe examine whether opioids differ in the riskof abuse and addiction. Although standardizeddefinitions for abuse and addiction have been
Vol. 26 No. 5 November 2003 1029Long-Acting Oral Opioids for Chronic Non-Cancer Pain
proposed, they have not been consistently uti-lized in studies investigating this outcome.8,9
We recorded any information about abuse andaddiction, including rates of death and hospital-ization when available.
Because of inconsistent reporting of out-comes, withdrawal rates may be a more reliablemeasure in studies of opioids. This outcome maybe a surrogate measure for either clinical effi-cacy or adverse events. One trial that examinedreason for withdrawal found different reasonsin its arms: withdrawals were due to adverseevents in patients on long-acting oxycodone, butdue to inadequate pain control in the patientson placebo.10 High withdrawal rates probablyindicate some combination of poor tolerabil-ity and ineffectiveness. An important subset iswithdrawal due to any adverse event (those whodiscontinue specifically because of adverseeffects).
Study Types. We included controlled clinicaltrials to evaluate efficacy. The validity of con-trolled trials depends on how they are designed.Randomized, properly blinded clinical trials areconsidered the highest level of evidence forassessing efficacy.11–13 Clinical trials that are notrandomized or blinded, and those that haveother methodologic flaws, are less reliable, butare also discussed in our report.
Trials that evaluated one long-acting opioidagainst another long-acting opioid provideddirect evidence of comparative efficacy andadverse event rates. Trials that compared long-acting opioids to short-acting opioids, non-opioids, or placebos provided indirectcomparative data.
To evaluate adverse event rates, we includedclinical trials and observational cohort studies.Clinical trials are often not designed to assessadverse events, and may select patients at low-risk for adverse events (in order to minimizedropout rates) and utilize methodology inad-equate for assessing adverse events. Observa-tional studies designed to assess adverse eventrates may include broader populations, carryout observations over a longer time period, uti-lize higher quality methodologic techniques forassessing adverse events, or examine largersample sizes.
One unique issue that complicates the inter-pretation of studies of chronic pain is “incom-plete cross-tolerance.” In medical jargon, a
patient who finds that a particular opioid isless effective over time is said to have become“tolerant” to that drug. “Incomplete cross-toler-ance” means that a patient’s “tolerance” for oneopioid may not carry over to other opioids. Ac-cording to the theory of incomplete cross-toler-ance, individuals who have been taking oneopioid may do better if they switch to a differentopioid (“opioid rotation”)—not because thenew one is a better drug, but simply because itis not the one they have been taking. In observa-tional studies of both cancer and non-cancerpatients, there is some evidence that incom-plete cross-tolerance occurs.14–17 In some cases,opioid rotation may be done to minimize ad-verse events.
MethodsLiterature Search
To identify articles relevant to each key ques-tion, we searched, in order, the Cochrane Li-brary (2002, Issue 1), MEDLINE (1966–2002),EMBASE (1980–2001), and reference lists ofreview articles. In electronic searches, we com-bined terms for pain with terms for opioid anal-gesics and narcotics, and relevant researchdesigns (see Appendix A for complete searchstrategy). In addition, the State of Oregon cre-ated and disseminated a submission protocolto pharmaceutical manufacturers for the sub-mission of clinical and economic evaluationdata to the Evidence-Based Practice Center. Allcitations were imported into an electronic data-base (EndNote 5.0). Searches on the electronicdatabases were carried out through October2002, using updates on electronic databasesafter the initial searches.
Study SelectionAll English-language titles and abstracts and
suggested additional citations were reviewed forinclusion using criteria developed by the re-search team with input from the subcommit-tee. We obtained full-text articles if the title andabstract review met the following eligibilitycriteria:
1. Systematic reviews of the clinical efficacyor adverse event rates of long-acting opi-oids in patients with chronic non-cancerpain, OR
1030 Vol. 26 No. 5 November 2003Chou et al.
2. Randomized controlled trials that com-pared one of the long-acting opioids listedabove to another long-acting opioid, ashort-acting opioid, a non-opioid, orplacebo in adult patients with chronicnon-cancer pain, OR
3. Randomized controlled trials and observa-tional studies that reported adverse eventrates for one of the long-acting opioidslisted above.
We re-applied these eligibility criteria to thefull-text articles, ensuring that the clinical effi-cacy or adverse event rates from specific opioidswere reported or could be calculated. Althoughstudies of longer duration were preferred, wehad no lower limit on the length of follow-up,but excluded “single-dose studies,” which exam-ine the effects of a single dose of medicationrather than a course of treatment.
Searches identified 3495 citations: 1081 fromthe Cochrane Library, 1106 from Medline, 1205from EMBASE, 42 from reference lists, and60 from pharmaceutical company submissions.We identified 1226 clinical trials and excluded1195 of these (see Appendix C for detailedsearch results). Nine hundred twenty-two clini-cal trials were excluded because they did notevaluate an included population (most ex-cluded studies evaluated patients with acutepain or cancer pain), 252 were excludedbecause they did not evaluate an included inter-vention (long-acting opioid), and 22 were ex-cluded because they did not evaluate anincluded outcome (pain control, pain relief,or function). Thirty-one trials were retrievedfor more detailed evaluation. After this sec-ond review, we excluded 14 trials: 10 becausethey did not evaluate an included interventionand 4 because they did not evaluate an includedpopulation. One additional randomized trialwas excluded because it used either long-actingmorphine or oxycodone in its opioid interven-tion group, and did not provide separate resultsfor each long-acting opioid.18 Sixteen random-ized controlled trials provided usable data andare included in evidence tables. We also re-viewed eight observational studies that evalu-ated adverse event rates from long-actingopioids.
Data AbstractionOne reviewer abstracted the following data
from included trials: study design, setting, pop-ulation characteristics (including sex, age, race,
diagnosis), eligibility and exclusion criteria, in-terventions (dose and duration), comparisons,numbers screened, eligible, enrolled, and lostto follow-up, method of outcome ascertain-ment (e.g., scales used), and results for eachoutcome. Although there is no clear consensuson “true” equianalgesic doses of opioid medica-tions, equianalgesic doses were estimated usingpublished tables.19 We recorded intention-to-treat results if available and the trial did notreport high overall loss to follow-up. In trialswith crossover, because of the potential for dif-ferential withdrawal prior to crossover biasingsubsequent results, outcomes for the first inter-vention were recorded if available. A secondreviewer checked all data.
Quality AssessmentWe assessed quality of trials based on the pre-
defined criteria listed in Appendix B. We ratedthe internal validity of each trial based on themethods used for randomization, allocationconcealment, and blinding; the similarity ofcompared groups at baseline; maintenance ofcomparable groups; adequate reporting of drop-outs, attrition, crossover, adherence, and con-tamination; loss to follow-up; and the use ofintention-to-treat analysis. External validity oftrials was assessed based on adequately describ-ing the study population, similarity of patientsto other populations to whom the interventionwould be applied, control group receiving com-parable treatment, funding source, and role ofthe funder.
Overall quality was assigned based on criteriadeveloped by the US Preventive Services TaskForce and the National Health Service Centerfor Reviews and Dissemination (UK).11,12 Trialsthat had a fatal flaw in one or more catego-ries were rated poor-quality; trials that met allcriteria were rated good-quality; the remainderwas rated fair-quality. As the “fair-quality” cate-gory is broad, studies with this rating vary intheir strengths and weaknesses: the results ofsome fair-quality studies are unlikely to be valid,while others are probably or likely to be valid. A“poor-quality” trial is not valid—the results areat least as likely to reflect flaws in the studydesign rather than true differences between thecompared drugs. A particular randomized trialmight receive two different ratings: one for effi-cacy and another for adverse events.
Vol. 26 No. 5 November 2003 1031Long-Acting Oral Opioids for Chronic Non-Cancer Pain
Appendix D shows the criteria we used torate clinical trials and observational studies ofadverse events. These criteria reflect aspectsof the study design that are particularly im-portant for assessing adverse event rates. Werated observational studies as good quality foradverse event assessment if they adequately metsix or more of the seven pre-defined criteria,fair if they met three to five criteria, and poorif they met two or fewer criteria.
After assignment of quality ratings by the ini-tial reviewer, quality ratings were independentlyassigned by a second reviewer. Overall qualityrating and quality rating scores (for studies onadverse event assessment) were compared be-tween reviewers. If overall quality ratings dif-fered, the two reviewers came to consensusprior to assigning a final quality rating.
Data SynthesisWe constructed evidence tables showing the
study characteristics, quality ratings, and resultsfor all included studies.
To assess the overall strength of evidence fora body of literature about a particular key ques-tion, we examined the consistency of study de-signs, patient populations, interventions, andresults. Consistent results from good-qualitystudies across a broad range of populationswould suggest a high degree of certainty thatthe results of the studies were true (that is, theentire body of evidence would be considered“good quality.”) For a body of fair-quality stud-ies, however, consistent results may indicate thatsimilar biases are operating in all the studies.
ResultsOverview of Included Trials
We identified 16 randomized trials (1427 pa-tients enrolled) that evaluated long-acting opi-oids in chronic non-cancer pain populations(Table 1). Recent non-systematic reviews on ad-verse events from opioids have identified onlytwo trials each.2,6 We did not find a relevantsystematic review for any of the key questions.
Only two of the 16 trials compared one long-acting opioid to another.20,21 One of thesetrials20 compared transdermal fentanyl to long-acting morphine; the other21 compared a once-daily morphine preparation to a twice-dailymorphine preparation. Seven trials compared a
long-acting opioid to a short-acting opioid,22–28
and seven compared a long-acting opioid to anon-opioid or placebo.10,29–34 Seven trials useda crossover design.20,24,25,29,31,32,34 We identi-fied trials on long-acting oxycodone,10,23,25,28,34
long-acting morphine,20,21,26,30–32 long-actingdihydrocodeine,24,27 long-acting codeine,22,29,33
and transdermal fentanyl.20 We did not identifyany trials on levorphanol or methadone. Onesmall trial35 with a high rate of withdrawal (14/20) cited in reference lists2,29 could not be lo-cated despite searches for journal, title, andauthor.
The trials ranged in size from 1231 to 29521
evaluable enrollees, with an average of 79 en-rollees. Five of the trials focused on osteoarthri-tis,10,21,23,27,33 five on back pain,22,24–26,28 two onneuropathic pain,30,34 one on phantom limbpain,31 and three on heterogeneous chronicnon-cancer pain.20,29,32
All of the trials were of relatively short dura-tion, ranging from 5 days22 to 16 weeks.26 Alltrials excluded persons with past or current sub-stance abuse. The majority of trials recruitedpatients from specialty clinics, most commonlyfrom rheumatology or pain practices. Race wasrarely reported. Sex had a slight predominance(slightly greater than 50%) towards women.The average age of enrollees was 55.
Assigned quality ratings did not differ be-tween reviewers.
Key Question Outcomes1A. In head-to-head comparisons, has one or
more long-acting opioid been shown to be supe-rior to other long-acting opioids in reducingpain and improving functional outcomes whenused for treatment of adults with refractorynon-cancer pain?
Two trials directly compared the efficacy ofone long-acting opioid to another in chronicpain of non-cancer origin.20,21 One trial20 com-pared transdermal fentanyl to long-acting mor-phine twice a day. The other trial21 compareda once-daily morphine preparation to a twice-daily morphine preparation.
The study that compared transdermal fen-tanyl to oral long-acting morphine used acrossover design in a population of 256 hetero-geneous chronic pain patients with an averageduration of 9 years pain.20 This study wasrated poor quality because of several major
1032 Vol. 26 No. 5 November 2003Chou et al.
Tab
le1
Ove
rvie
wof
All
Lon
g-A
ctin
gO
pioi
dT
rial
s
Ave
rage
Pain
Sam
ple
Dos
eIn
ten
sity
Ave
rage
Res
cue
Aut
hor
,Ye
arL
ong-
Act
ing
Opi
oid
Stud
yT
ype
Pain
Typ
eD
urat
ion
Size
(mg/
day)
Scor
eSc
ale
Res
cue
Dru
gD
rug
Usa
ge
Lon
g-ac
ting
vs.
long
-act
ing
tria
lsA
llan
2001
A:
Tra
nsd
erm
alC
ross
over
Mis
cella
neo
us4
wee
ksa
212
NR
A:
57.8
0–10
0V
AS
Not
spec
ifie
d29
.4m
g/da
yfe
nta
nyl
B:
Ora
lm
orph
ine
B:
62.9
23.6
mg/
day
(tw
ice
daily
)C
aldw
ell
2002
A:
Mor
phin
eR
CT
Ost
eoar
thri
tis
4w
eeks
295
A:
30m
gA
:31
30–
500
VA
SN
otpe
rmit
ted
N/A
(on
ceda
ilya.
m.)
B:
Mor
phin
eB
:30
mg
B:
326
(on
ceda
ilyp.
m.)
C:
Mor
phin
eC
:30
mg
C:
322
(tw
ice
daily
)L
ong-
actin
gvs
.pl
aceb
oor
non-
opio
idtr
ials
Ark
inst
all
1995
Cod
ein
e(v
s.pl
aceb
o)C
ross
over
Mis
cella
neo
us7
days
a46
353
350–
100
VA
ST
ylen
olw
ith
3.6
tabs
/day
code
ine
Pelo
so20
00C
odei
ne
(vs.
plac
ebo)
RC
TO
steo
arth
riti
s4
wee
ks10
315
932
.50–
100
VA
ST
ylen
ol4.
2ta
bs/d
ayH
arke
2001
Mor
phin
e(v
s.ca
rbam
azep
ine)
RC
TN
euro
path
ic8
days
3883
6.9b
0–10
VA
SN
otpe
rmit
ted
N/A
pain
Hus
e20
01M
orph
ine
(vs.
plac
ebo)
Cro
ssov
erPh
anto
m4
wee
ksa
1211
53.
620–
10V
AS
Asp
irin
�N
Rlim
bpa
inpa
race
tam
olM
oulin
1996
Mor
phin
e(v
s.be
nzt
ropi
ne)
Cro
ssov
erM
isce
llan
eous
6w
eeks
a61
83.4
450–
100
VA
SPa
race
tam
ol3.
5ta
bs/d
ayR
oth
2000
Oxy
codo
ne
(vs.
plac
ebo)
RC
TO
steo
arth
riti
s2
wee
ks13
340
1.6
0–3
Cat
.N
otpe
rmit
ted
N/A
Wat
son
1998
Oxy
codo
ne
(vs.
plac
ebo)
Cro
ssov
erPo
sth
erpe
tic
4w
eeks
a50
4535
0–10
0V
AS
Not
perm
itte
dN
/An
eura
lgia
Lon
g-ac
ting
vs.
shor
t-act
ing
tria
lsC
aldw
ell
1999
Oxy
codo
ne
RC
TO
steo
arth
riti
s4
wee
ks10
740
1.3
0–4
Cat
.N
otpe
rmit
ted
N/A
(vs.
SAox
ycod
one
�ac
etam
inop
hen
)H
ale
1999
Oxy
codo
ne
Cro
ssov
erB
ack
pain
6da
ysa
4740
1.2
0–3
Cat
.SA
Oxy
codo
ne
0.6
tabs
/day
(vs.
SAox
ycod
one)
5–10
mg
PRN
Salz
man
1998
Oxy
codo
ne
RC
TB
ack
pain
10da
ys57
401.
10–
3C
at.
SAO
xyco
don
eN
R(v
s.SA
oxyc
odon
e)5–
10m
gPR
NH
ale
1997
Cod
ein
eR
CT
Bac
kpa
in5
days
8320
01.
60–
4C
at.
Ace
tam
inop
hen
4.0
tabs
/day
(vs.
acte
min
oph
en�
SAco
dein
e)G
osti
ck19
89D
ihyd
roco
dein
eC
ross
over
Bac
kpa
in2
wee
ksa
61N
R1.
75N
otpr
ovid
edPa
race
tam
ol1.
54ta
bs/d
ay(v
s.SA
dih
ydro
code
ine)
Llo
yd19
92D
ihyd
roco
dein
eR
CT
Ost
eoar
thri
tis
2w
eeks
86N
R39
.20–
100
VA
SN
otpe
rmit
ted
N/A
(vs.
dext
ropr
opox
yph
ene
�pa
race
tam
ol)
Jam
ison
1998
LA
Mor
phin
e�
SAR
CT
Bac
kpa
in16
wee
ks36
4154
.90–
100
VA
SN
otpe
rmit
ted
N/A
oxyc
odon
e(v
s.SA
oxyc
odon
e)a D
urat
ion
per
inte
rven
tion
ofcr
osso
ver
tria
l.b M
axim
umpa
inin
ten
sity
prio
rto
reac
tiva
tion
ofsp
inal
cord
stim
ulat
ion
unit
.R
CT
�ra
ndo
miz
edco
ntr
olle
dtr
ial;
VA
S�
visu
alan
alog
uesc
ale;
PRN
�as
nee
ded;
SA�
shor
t-act
ing
opio
idpr
epar
atio
n;N
R�
not
repo
rted
.
Vol. 26 No. 5 November 2003 1033Long-Acting Oral Opioids for Chronic Non-Cancer Pain
methodologic flaws (Evidence Table 1.1). Themost important areas of concern were that nei-ther patients nor investigators were blinded,and many of the trial participants were on oneof the study drugs prior to entry. Blinding isparticularly important in studies using subjec-tive measures. This may have been an evengreater factor in this trial, in which 76% of theenrollees were taking morphine prior to enroll-ment. Patients who had achieved better resultswith morphine were probably less likely toenroll. If subjects who were entered into the trialhad responded poorly to morphine relative toother patients, they could have been favorablypredisposed towards a new medication. Incom-plete cross-tolerance could also have biased theresults towards transdermal fentanyl simplybecause it was new.
This study found that, after 4 weeks of treat-ment, more patients reported good or verygood pain control for fentanyl (40%) than formorphine (19%). On the other hand, with-drawal rates favored long-acting morphine(9%) over fentanyl (16%). Functional out-comes were assessed using SF-36 and favoredfentanyl, though raw numbers were not re-ported. A subgroup analysis of the 66 enrolleeswho were naı̈ve to morphine and fentanyl atthe beginning of the study found equivalentwithdrawal rates between interventions.
How similar was the study sample to the popu-lation of interest to the clinician? As discussedabove, the subjects can best be described aspatients who have not had a good response tomorphine or another opioid in the first place.The question it appears to address is, “do pa-tients with chronic non-cancer pain accus-tomed to opioids (and who may not have had agood response to morphine or another opioid inthe first place) prefer a change to transdermalfentanyl?” The study does not address the ques-tion of greater interest to clinicians: “in unse-lected patients who have chronic pain requiringtreatment with opioids, is transdermal fentanylmore effective than long-acting morphine?”
Other aspects of the trial make its externalvalidity difficult to assess. Exclusion criteriawere not specified, and the numbers of patientsscreened and eligible for entry were not re-ported. Patients in both groups took immedi-ate-release morphine as needed to supplementtheir long-acting medication. The length of
follow-up for each drug regimen was onlyfour weeks.
The study that compared a once-daily mor-phine preparation to a twice-daily morphinepreparation21 used a randomized, double-blinded design in a population of 295 osteoar-thritis patients. Four treatment groups wereevaluated: once-daily morphine in the morn-ing, once-daily morphine in the evening, twice-daily morphine, and placebo. This study wasrated fair quality and appeared to use ade-quate blinding and randomization (EvidenceTable 1.1). Important limitations included noevaluation of the blinding, no comparison ofpersons who completed the study, high over-all withdrawal rates, and no explanation ofhow withdrawn patients were handled in dataanalysis.
This study found that once-daily morphinewas not significantly different than twice-dailymorphine for all measures of pain control (Evi-dence Table 1.1). For sleep, one of seven mea-sures of sleep quality (overall sleep quality)showed a slight but significant improvement inpatients receiving once-daily morphine in themorning (but not once-daily morphine in theevening) compared to twice-daily morphine; allother measures of sleep quality were not signifi-cantly different between once- and twice-dailymorphine. All three morphine treatmentgroups were better than placebo for most mea-sures of efficacy. Withdrawal rates were similarin all active treatment groups.
External validity of this trial was difficult toassess because the numbers of patients screenedand eligible for entry were not reported, thelength of follow-up for each drug regimen wasonly four weeks, and duration of pain and previ-ous opioid use in evaluated patients was notreported.
We found no trials directly comparing fen-tanyl or long-acting morphine to any otherlong-acting preparations.
1B. In trials comparing long-acting opioidsto other types of drugs or to placebo, is therea pattern to suggest that one long-acting opioidis more effective than another?
We identified 14 fair-quality trials (876 pa-tient enrolled) that gave indirect evidence re-garding the comparative efficacy of long-actingopioids. These trials exhibited a high degreeof heterogeneity with respect to study designs,
1034 Vol. 26 No. 5 November 2003Chou et al.
patient populations, interventions, and out-comes measured (Table 1). All studies wererated fair quality (see Evidence Tables 1.2 and1.3) and had at least one of the following meth-odologic problems: inadequate or poorlydescribed randomization and allocation con-cealment, lack of blinding or unclear blindingmethods, or high loss to follow-up.
Three trials evaluated long-acting codeine,22,
29,33 two long-acting dihydrocodeine,24,27 fourlong-acting morphine,26,30–32 and five long-acting oxycodone.10,23,25,28,34 The averageequipotent opioid dose received varied greatlyand in two trials was not reported.24,27 The dura-tion of follow-up ranged from 5 days to 16weeks, and a wide range of outcomes and mea-sures were employed. The most common out-comes assessed were pain intensity and rescuedrug use (Table 1). The studies used differentpain intensity measures, the most commonbeing VAS.
For most outcomes of clinical efficacy, thescales used varied too much across trials to drawmeaningful comparisons between differentlong-acting opioids. For pain intensity, for ex-ample, of five trials on oxycodone, one used aVAS,34 three others used two different (0–310,25
or 0–423) categorical scales, and one did notreport pain intensity as an outcome.28 For theoutcomes pain intensity, pain relief, and func-tional outcome, there did not appear to be apattern favoring one long-acting opioid overanother.
Functional outcomes assessment also variedwidely between studies. For sleep, the mostwidely reported functional outcome, measure-ment tools used were sleep quality (1–5 scale23
or 0–10 scale,10) nighttime rescue medicationuse,22 hours of sleep,26 average nights awakenedby pain,27 and VAS (1–100) for trouble fallingasleep and needing medication to sleep.33 Be-cause of the heterogeneity of scales used tomeasure sleep quality, meaningful comparisonsbetween long-acting opioids could not be made.Other functional outcomes were less commonlyreported and when reported were also character-ized by marked heterogeneity in measurementscales. We were unable to perform meta-analysison any subgroup of trials.
Withdrawal rates were reported in all studiesand also did not exhibit a pattern favoring onelong-acting opioid versus other long-acting opi-oids (Table 2). For long-acting oxycodone, the
withdrawal rate ranged from 4%25 to 53%.10
For long-acting morphine, the withdrawal rateranged from 0%31 to 30%.32 Similar wide rangesfor withdrawal rates were seen for the studies onlong-acting dihydrocodeine and long-acting co-deine. The wide range of withdrawal rates couldreflect differences in populations, dosing ofmedications in trials, use of a run-in period, orother factors.
The trials generally provided inadequate in-formation to accurately assess external validityor showed evidence of having highly selectedpopulations. Most trials did not report numbersof patients screened or eligible for entry andsome did not specify exclusion criteria. Whenexclusion criteria were specified, patients at riskfor drug or substance abuse were typically ex-cluded from trial participation. Numbersexcluded for meeting specific exclusion criteriawere usually not reported.
1C. Have long-acting opioids been shown tobe superior to short-acting opioids in reducingpain and improving functional outcomes whenused for treatment of adults with chronicnon-cancer pain?
A subgroup of 7 (568 patients enrolled) ofthe 14 trials reviewed for key question 1B di-rectly compared the efficacy of long-actingopioids to short-acting opioids in patients withchronic non-cancer pain (Table 3). All wererated fair quality (Evidence Table 1.2). Threestudies compared long-acting to short-acting ox-ycodone.23,25,28 One of these studies25 re-ran-domized patients who had enrolled in a previoustrial.28 Two studies evaluated long-acting di-hydrocodeine,24,27 one evaluated long-actingcodeine,22 and one evaluated long-acting mor-phine.26 Study designs, patient populations, andoutcomes assessed varied between studies (Evi-dence Table 1.2).
These trials showed no consistent trendsdemonstrating significant differences in effi-cacy between long-acting opioids as a class andshort-acting opioids (Table 3). Three studiesthat found differences in efficacy favoring long-acting morphine,26 long-acting dihydrocod-eine,27 and long-acting codeine22 had featuresthat might invalidate these results. In thetrials on long-acting morphine26 and long-acting codeine,22 the average daily doses ofopioid in the long-acting arm were higher thanthe average daily doses given in the short-acting
Vol. 26 No. 5 November 2003 1035Long-Acting Oral Opioids for Chronic Non-Cancer Pain
Tab
le2
Wit
hdra
wal
Rat
es
Ove
rall
Inad
equa
teA
uth
or,
Year
Lon
g-A
ctin
gO
pioi
dW
ith
draw
alR
ates
Wit
hdr
awal
Rat
esPe
rD
rug
Pain
Con
trol
Adv
erse
Eff
ects
Oth
er
Lon
g-ac
ting
vs.
long
-act
ing
tria
lsA
llan
2001
A:
Fen
tan
yltr
ansd
erm
al23
%A
:15
%(3
8/25
0)N
/A“1
1%”
N/A
B:
Mor
phin
eor
al(t
wic
eda
ily)
B:
9%(2
2/23
8)“4
%”
Cal
dwel
l20
02A
:M
orph
ine
(on
ceda
ilya.
m.)
38%
A:
37%
(27/
73)
917
1B
:M
orph
ine
(on
ceda
ilyp.
m.)
B:
45%
(33/
73)
1218
3C
:M
orph
ine
(tw
ice
daily
)C
:37
%(2
8/76
)8
182
D:
Plac
ebo
D:
32%
(23/
72)
145
4L
ong-
actin
gvs
.pl
aceb
oor
non-
opio
idtr
ials
Ark
inst
all
1995
Cod
ein
e28
%L
AC
odei
ne:
19%
(9/4
6)1
71
Plac
ebo:
9%(4
/46)
01
3Pe
loso
2000
Cod
ein
e36
%L
AC
odei
ne:
40%
(20/
51)
115
4Pl
aceb
o:33
%(1
7/52
)5
57
Har
ke20
01M
orph
ine
8%L
AM
orph
ine:
5%(1
/19)
NR
NR
1Pl
aceb
o:11
%(2
/19)
2H
use
2001
Mor
phin
e0%
LA
Mor
phin
e:0%
(0/1
2)N
RN
/AN
/APl
aceb
o:0%
(0/1
2)M
oulin
1996
Mor
phin
e30
%L
AM
orph
ine:
NR
NR
NR
Rot
h20
00O
xyco
don
e53
%L
AO
xyco
don
e20
mg:
42%
(19/
44)
514
0L
AO
xyco
don
e10
mg:
50%
(24/
44)
1212
0Pl
aceb
o:60
%(2
7/45
)22
23
Wat
son
1998
Oxy
codo
ne
22%
LA
Oxy
codo
ne:
12%
(6/5
0)0
51
Plac
ebo:
10%
(5/5
0)1
31
Lon
g-ac
ting
vs.
shor
t-act
ing
tria
lsC
aldw
ell
1999
Oxy
codo
ne
34%
LA
Oxy
codo
ne:
21%
(7/3
4)3
31
SAO
xyco
don
e:30
%(1
1/37
)4
52
Plac
ebo:
50%
(18/
36)
133
2H
ale
1999
Oxy
codo
ne
6%L
AO
xyco
don
e:4%
(2/4
7)0
20
SAO
xyco
don
e:2%
(1/4
7)0
11
Salz
man
1998
Oxy
codo
ne
18%
LA
Oxy
codo
ne:
20%
(6/3
0)N
R6
NR
SAO
xyco
don
e:7%
(2/2
7)0
20
(on
lyad
vers
eev
ent
wit
hdr
awal
sre
port
ed)
Hal
e19
92C
odei
ne
22%
LA
Cod
ein
e:32
%(1
7/53
)1
151
SAC
odei
ne:
12%
(6/5
1)1
50
Gos
tick
1989
Dih
ydro
code
ine
26%
NR
NR
NR
NR
Llo
yd19
92D
ihyd
roco
dein
e34
%L
AD
ihyd
roco
dein
e:47
%(2
0/43
)1
172
Dex
trop
ropo
xyph
ene
�pa
race
tam
ol:
21%
(9/4
3)2
43
Jam
ison
1998
Mor
phin
e8%
LA
Mor
phin
e�
SAO
xy.:
6%(1
/18)
NR
1N
RSA
Oxy
codo
ne:
12%
(2/1
8)0
20
1036 Vol. 26 No. 5 November 2003Chou et al.
Table 3Overview of Randomized Controlled Trials of Long-Acting vs. Short-Acting Opioidsa
Author, Year Pain Type Duration Patients Findings
OxycodoneCaldwell 1999 Osteoarthritis 30 days 107 LA oxycodone and SA oxycodone plus Tylenol are equally effective
for pain control and improvement of sleep.Hale 1999 Back pain 6 daysb 47 LA oxycodone and SA oxycodone are equally effective
for pain control.Salzman 1998 Back pain 10 days 57 LA oxycodone and SA oxycodone are equally effective when titrated
for pain control.Codeine
Hale 1992 Back pain 5 days 83 LA codeine plus acetaminophen are more effective for pain controlthan SA codeine plus acetaminophen, but these drugs were notgiven at therapeutically equivalent doses.
DihydrocodeineGostick 1989 Back pain 2 weeksb 61 LA dihydrocodeine and SA dihydrocodeine are equally effective
for pain control.Lloyd 1992 Osteoarthritis 2 weeks 86 LA dihydrocodeine and SA dihydrocodeine are equally effective
for pain control when compared directly.Morphine
Jamison 1998 Back pain 16 weeks 36 LA morphine plus SA oxycodone together are more effectivefor pain control than SA oxycodone, but these drugs werenot given at therapeutically equivalent doses.
aAll trials are of fair quality.bDuration per intervention of crossover trial.LA � long-acting opioid preparation; SA � short-acting opioid preparation.
group. In the other study,27 significant differ-ences in pain relief were only seen when thelong-acting dihydrocodeine group was com-pared to itself at different points in time, butno significant differences were found when thelong-acting opioid was compared directly tothe short-acting opioid. Functional outcomeswere inconsistently examined or used hetero-geneous measurement scales. Other importantoutcomes such as improved compliance or moreconsistent pain control were not examined.
A subgroup of three trials of 281 enrolledpatients evaluated roughly equivalent doses oflong- and short-acting oxycodone and appearedto be the most homogeneous of this groupof trials.23,25,28 One of these trials25 investigated are-randomized population of patients studiedin a previous trial28 but used a different inter-vention protocol. These three trials found nosignificant differences in efficacy (pain relief)between long- and short-acting oxycodone.With regard to functional outcomes, one ofthese trials23 reported improved sleep qualitywith long-acting oxycodone, but baseline sleepscores were significantly better in patients ran-domized to this intervention, which could inval-idate this finding.
2. What are the comparative incidence andnature of adverse effects (including addictionand abuse) of long-acting opioid medications
in adult patients being treated for chronic non-cancer pain?
A variety of long-acting opioids are used fortreatment of chronic non-cancer pain. Therecontinue to be concerns, however, regardingthe risk of adverse events.9 Common adverseevents associated with opioid use includenausea, cognitive dysfunction, and constipa-tion. More serious but less common adverseevents include respiratory depression, abuse,and addiction. In non-cancer pain patients,data are lacking regarding differential risks forlong-acting opioids.6
2A. In head-to-head comparisons, has one ormore long-acting opioid been shown to be asso-ciated with fewer adverse events compared toother long-acting opioids when used for treat-ment of adults with chronic non-cancer pain?
As discussed earlier, only two randomizedtrials directly compared two long-acting opi-oids. One of these trials20 compared two differ-ent long-acting opioids (transdermal fentanyland long-acting morphine) and the other21
compared once-daily versus twice-daily prepara-tions of oral morphine. Neither study assessedrates of addiction or abuse. No deaths werereported in either study.
The trial which compared transdermal fen-tanyl with long-acting oral morphine was rated
Vol. 26 No. 5 November 2003 1037Long-Acting Oral Opioids for Chronic Non-Cancer Pain
poor-quality for adverse event assessment (Evi-dence Table 2.1).20 This trial failed to ade-quately meet 6 out of the 7 predefined criteriafor adverse event assessment. This trial foundno significant differences in reported rates ofoverall or “serious” (not defined) complica-tions. Constipation was significantly lower fortransdermal fentanyl compared to long-actingmorphine (29% vs. 48%, P � 0.001) only asassessed by a bowel function questionnaire, andnot by patient-reported or investigator-ob-served symptoms. The rate of withdrawals dueto adverse event for all patients favored long-acting oral morphine (11% vs. 4%, P value notreported), but did not differ significantly inthe subgroup not previously on fentanyl ormorphine.
The trial which compared once-daily versustwice-daily preparations of oral morphine wasalso rated poor quality for adverse events (Evi-dence Table 2.1).21 This trial failed to ade-quately meet 5 out of the 7 predefined criteriafor adverse event assessment. Serious compli-cations (not defined) occurred in 6 enrolledpatients, but the rates of serious complicationswere not reported for each treatment group.This trial found a significantly higher rate ofconstipation in patients on once-daily morphinegiven in the morning (49%) versus twice-dailymorphine (29%), but a lower rate of asthenia(1% vs. 9%). The overall withdrawal rates intreated patients were not significantly differentbetween interventions and ranged from 37–45%, with withdrawal rates due to adverseevents ranging from 23–25%.
2B. In trials comparing long-acting opioidsto other types of drugs or to placebo, is therea pattern to suggest that one long-acting opioidis associated with fewer adverse events thananother?
Randomized Trials. Of the 14 trials reviewed forkey question 1B, 13 (994 patients enrolled) re-ported adverse event rates from long-acting opi-oids in patients with chronic non-cancer pain.One trial of long-acting morphine versus carba-mazepine for neuropathic pain30 was excludedbecause accurate adverse event rates could notbe abstracted from the graphs in the article.
All 13 trials were rated fair or poor qualityfor adverse event assessment and had at leastimportant methodologic flaws (Table 4; Evi-dence Tables 2.2 and 2.3). In addition, these
trials had heterogeneous study designs, inter-ventions, outcomes, and patient populations,making meaningful comparisons across stud-ies difficult (Table 1). None of these trials as-sessed rates of abuse or addiction, and allexcluded patients at risk for these complications.
These trials reported wide ranges for adverseevent rates even in studies that evaluated thesame long-acting opioid at roughly equivalentdoses. For long-acting oxycodone at mean dosesof 40 mg, for example, rates of nausea rangedfrom 15%23 to 50%28 in 5 trials (Table 4). With-drawal rates due to adverse events ranged from4%25 to 32%10 in these same studies. Given theuncertainty regarding the adverse event ratesfor individual long-acting opioids, it is not sur-prising that these trials show no discernible pat-tern of one long-acting opioid being superior toothers for any reported adverse event (Table 4).
Observational Studies. We identified 8 observa-tional studies evaluating the risk of long-actingopioids in 1190 patients with non-cancerpain.10,21,29,36–40 All were rated poor quality foradverse event assessment except one,10 whichwas rated fair (Evidence Table 2.4). For 6 ofthe 8 studies, independently assigned qualityrating scores were identical between two review-ers. For two studies, quality rating scores dif-fered by one41 or two10 points; in neither casedid this difference result in a change in overallquality rating. The single study rated fairquality10 met only 4 out of 7 predefined qualityassessment criteria. The most important areasof concern in this study were high overall loss tofollow-up (60/106) and the failure to specifyor define adverse events in advance.
Opioids assessed were long-acting codeine,29
long-acting morphine (once-daily21 or twice-daily40), transdermal fentanyl,36,39 methadone,38
and long-acting oxycodone.10 One study assessedboth methadone and long-acting morphine.37
The number of patients on long-acting opioids inthese studies ranged from 1138 to 530.39 Fivewere prospective cohort studies10,21,29,36,39 andthree were retrospective cohorts.37,38,40 No iden-tified study was population-based. Three of theprospective cohorts10,21,29 were open-label ex-tensions of clinical trials included in this review.
Results of the observational studies were notsignificantly different from those reported inclinical trials for gastrointestinal adverse events,neurological adverse events, and withdrawal
1038 Vol. 26 No. 5 November 2003Chou et al.
rates due to adverse events (Table 5). The studyrated fair quality,10 for example, reported a rateof 31% (32/103) of withdrawal due to ad-verse events, which fell within the range fortrials of long-acting oxycodone.
Some observational studies reported long-term outcomes and serious adverse events notreported in the trials. The largest (n � 530)study39 reported one death (0.2%, 1/530)thought related to medication, four cases ofrespiratory depression (1%), and three epi-sodes of drug abuse (0.6%). Two other studiesreported rates of abuse,37,38 but they were retro-spective studies with small samples (n � 11 and20) and no inception cohort. Four studies re-ported rates of long-term use, which could bea long-term measure of tolerability or clinicalefficacy.10,21,29,36 Rates ranged from 19% fortransdermal fentanyl36 to 54% for long-actingcodeine.29
The patients enrolled did not appear to beless selected than those in the controlled trials.In the prospective cohort studies, at least someparticipants were recruited from completedclinical trials,10,21,29,36,39 resulting in an evenmore highly selected population than the origi-nal trials. In the retrospective studies, no incep-tion cohort was identified and the populationappeared to represent a “convenience” sampleof patients for whom data was readilyavailable.37,38,40
No meaningful conclusions regarding com-parative adverse event risk of long-acting opioidscan be drawn from these observational studies.
2C. Have long-acting opioids been shown tohave fewer adverse events than short-acting opi-oids when used for treatment of adults withchronic non-cancer pain?
A subgroup of 7 of the 13 randomized trialsreviewed for key question 2B directly comparedlong-acting with short-acting opioids.22–28
In the single trial in this group rated fairquality,26 adverse events were not prespecifiedor defined and patients and investigators werenot blinded. Furthermore, patients in one armof this trial were given higher doses of opioidsthan the other. Adverse events would be ex-pected to be more common in the group receiv-ing higher doses, the result observed for mostreported adverse events (Table 4).
Across all trials, no pattern favoring eitherlong-acting or short-acting opioids was evidentfor any of the reported adverse events (Table
6). In the three most comparable studies, whichinvestigated roughly equivalent daily doses ofoxycodone in short-acting and long-actingpreparations,23,25,28 no trends favoring one for-mulation over the other were seen for the out-comes of dizziness, somnolence, vomiting, andconstipation. This was also true in the twostudies25,28 that investigated the same (re-ran-domized) population.
Withdrawals due to adverse events were re-ported in five trials (Table 4). Three favoredshort-acting opioids,22,27,28 one favored long-acting,23 and one was equivocal.25 These data arelimited by the poor quality of the trials foradverse event assessment and the fact that twoof the trials evaluated the same population.
In summary, there is no convincing evidenceto suggest superior adverse event rates withlong-acting opioids as a class compared to short-acting opioids.
3. Are there subpopulations of patients (spe-cifically by race, age, sex, or type of pain) withchronic non-cancer pain for which one long-acting opioid is more effective or associatedwith fewer adverse effects?
No clinical trials or observational studies weredesigned to compare the efficacy of long-actingopioids for different races, age groups, or sexes.There is almost no information regarding thecomparative efficacy of long-acting opioids forspecific subpopulations as characterized by race,sex, or age. Race was rarely reported in the trials;when it was reported the overwhelming majorityof patients were white. Women were well repre-sented in the trials (slightly over 50%), but differ-ential efficacy or adverse event rates accordingto sex were not evaluated. The average age ofincluded patients was 55 years, and one study34
evaluated patients with an average age of 70years. Two trials10,23 performed very limited sub-group analysis on older patients; neither trial wasa direct comparison of one long-acting opioidversus another and provide little informationregarding differential efficacy or adverse eventswithin the class of long-acting opioids.
Several specific types of chronic non-cancerpain patients were studied in some of the re-viewed trials. These categories included backpain,22,24–26,28 osteoarthritis,10,23,27,33 phantomlimb pain,31 neuropathic pain,30 and posther-petic neuralgia.34 None of these trials are direct
Vol. 26 No. 5 November 2003 1039Long-Acting Oral Opioids for Chronic Non-Cancer Pain
Tab
le4
Stud
yC
hara
cter
isti
csan
dA
dver
seE
vent
s,T
rial
sof
Lon
g-A
ctin
gO
pioi
ds
Con
fusi
onor
Qua
lity
Dro
wsi
nes
sor
Dif
ficu
lty
Stud
yIn
terv
enti
ons
Rat
inga
Nau
sea
Vom
itin
gC
onst
ipat
ion
Som
nol
ence
Diz
zin
ess
Con
cen
trat
ing
Wit
hdr
awal
b
Lon
g-ac
ting
oxyc
odon
eC
aldw
ell
1999
A:
Lon
g-ac
tin
gPO
OR
(2)
A:
15%
(5/3
4)A
:6%
(2/3
4)A
:71
%(2
4/34
)A
:53
%(1
8/34
)A
:12
%(4
/34)
Not
repo
rted
A:
6%(3
/34)
oxyc
odon
eB
:Sh
ort-a
ctin
gB
:38
%(1
4/37
)B
:11
%(4
/37)
B:
54%
(20/
37)
B:
70%
(26/
37)
B:
24%
(9/3
7)B
:14
%(5
/37)
oxyc
odon
e�
acet
amin
oph
enH
ale
1999
A:
Lon
g-ac
tin
gPO
OR
(2)
A:
16%
(4/2
5)A
:0%
(0/2
5)A
:32
%(8
/25)
A:
12%
(3/2
5)A
:16
%(4
/25)
Not
repo
rted
A:
4%(2
/47)
oxyc
odon
eB
:Sh
ort-a
ctin
gB
:41
%(9
/22)
B:
0%(0
/22)
B:
45%
(10/
22)
B:
18%
(4/2
2)B
:9%
(2/2
2)B
:2%
(1/4
7)ox
ycod
one
Rot
h20
00A
1:L
ong-
acti
ng
FAIR
(4)
A1:
41%
(18/
44)
A1:
23%
(10/
44)
A1:
32%
(14/
44)
A1:
27%
(12/
44)
A1:
20%
(9/4
4)N
otre
port
edA
1:32
%(1
4/44
)ox
ycod
one
20m
gbi
dA
2:L
ong-
acti
ng
A2:
27%
(12/
44)
A2:
11%
(5/4
4)A
2:23
%(1
0/44
)A
2:25
%(1
1/44
)A
2:30
%(1
3/44
)cA
2:27
%(1
2/44
)ox
ycod
one
10m
gbi
dB
:Pla
cebo
B:
11%
(5/4
5)B
:7%
(3/4
5)B
:7%
(3/4
5)B
:4%
(2/4
5)B
:9%
(4/4
5)B
:4%
(2/4
5)Sa
lzm
an19
98A
:L
ong-
acti
ng
POO
R(2
)A
:50
%(1
5/30
)A
:20
%(6
/30)
A:
30%
(9/3
0)A
:27
%(8
/30)
A:
30%
(9/3
0)A
:3%
(1/3
0)A
:20
%(6
/30)
oxyc
odon
eB
:Sh
ort-a
ctin
gB
:33
%(9
/27)
B:
4%(1
/27)
B:
37%
(10/
27)
B:
37%
(10/
27)
B:
22%
(6/2
7)B
:0%
(0/2
7)B
:7%
(2/2
7)ox
ycod
one
Wat
son
1998
A:
Lon
g-ac
tin
gFA
IR(3
)N
otre
port
edN
otre
port
edN
otre
port
edN
otre
port
edN
otre
port
edN
otre
port
edN
otre
port
edox
ycod
one
B:P
lace
boL
ong-
actin
gco
dein
eA
rkin
stal
ldA
:L
ong-
acti
ng
FAIR
(3)
A:
33%
cA
:14
%A
:21
%A
:16
%A
:21
%N
otre
port
edA
:15
%(7
/46)
1995
code
ine
B:P
lace
boB
:12%
B:
3.8%
B:
10%
B:
5%B
:14%
B:
2%(1
/46)
Hal
e19
97A
:L
ong-
acti
ng
POO
R(1
)A
:31
%(1
6/52
)A
:10
%(5
/52)
A:
19%
(10/
52)
A:
10%
(5/5
2)A
:17
%(9
/52)
Not
repo
rted
A:
25%
(13/
53)
code
ine
B:S
hor
t-act
ing
B:
18%
(9/5
1)B
:2%
(1/5
1)B
:16
%(8
/51)
B:
4%(2
/51)
B:
4%(2
/51)
B:
8%(4
/51)
code
ine
(con
tinue
d)
1040 Vol. 26 No. 5 November 2003Chou et al.
Tab
le4
Con
tinu
ed
Con
fusi
onor
Qua
lity
Dro
wsi
nes
sor
Dif
ficu
lty
Stud
yIn
terv
enti
ons
Rat
inga
Nau
sea
Vom
itin
gC
onst
ipat
ion
Som
nol
ence
Diz
zin
ess
Con
cen
trat
ing
Wit
hdr
awal
b
Pelo
so20
00A
:L
ong-
acti
ng
FAIR
(3)
Not
repo
rted
Not
repo
rted
A:
49%
(25/
51)c
A:
39%
(20/
51)
A:
33%
(17/
51)
Not
repo
rted
A:
29%
(15/
51)
code
ine
B:P
lace
boB
:11
%(6
/52)
B:
10%
(5/5
2)B
:8%
(4/5
2)B
:8%
(4/5
2)L
ong-
actin
gdi
hydr
ocod
eine
Gos
tick
1989
A:
Lon
g-ac
tin
gPO
OR
(2)
Not
repo
rted
Not
repo
rted
A:
55%
(23/
42)e
Not
repo
rted
Not
repo
rted
Not
repo
rted
Not
repo
rted
dih
ydro
code
ine
B:S
hor
t-act
ing
B:
49%
(21/
43)
dih
ydro
code
ine
Llo
ydf
1992
A:
Lon
g-ac
tin
gPO
OR
(2)
A:
31%
(12/
39)
Not
repo
rted
A:
8%(3
/39)
A:
26%
(10/
39)
Not
repo
rted
A:
10%
(4/3
9)A
:40
%(1
7/43
)di
hyd
roco
dein
eB
:B
:10
%(4
/41)
B:
10%
(4/4
1)B
:15
%(6
/41)
Not
repo
rted
B:5
%(2
/41)
B:
9%(4
/43)
Dex
trop
ropo
xyph
ene
�pa
race
tam
olL
ong-
actin
gm
orph
ine
Hus
eg20
01A
:L
ong-
acti
ng
FAIR
(3)
A:
0.74
cmN
otre
port
edA
:0.
03cm
cA
:2.
21cm
A:
1.27
cmN
otre
port
edN
otre
port
edm
orph
ine
B:P
lace
boB
:0.
4cm
B:
0.02
cmB
:1.
33cm
B:
0.71
cmJa
mis
ond
1998
A:
Lon
g-ac
tin
gFA
IR(5
)A
:31
%N
otre
port
edA
:30
%A
:31
%A
:6%
A:
1%N
otre
port
edm
orph
ine
�sh
ort-a
ctin
gox
ycod
one
B:S
hor
t-act
ing
B:
14%
B:
18%
B:
14%
B:
19%
B:
1.4%
oxyc
odon
eM
oulin
1996
A:
Lon
g-ac
tin
gFA
IR(4
)A
:39
%(1
8/46
)cA
:39
%(1
8/46
)cA
:41
%(1
9/46
)cN
otre
port
edA
:37
%(1
7/46
)A
:9%
(4/4
6)A
:28
%h
(13/
46)
mor
phin
eB
:Ben
ztro
pin
eB
:7%
(3/4
6)B
:2%
(1/4
6)B
:4%
(2/4
6)B
:2%
(1/4
6)B
:15
%(7
/46)
B:
2%(1
/46)
a Num
ber
ofcr
iter
iaou
tof
seve
nad
equa
tely
met
.b D
ueto
adve
rse
even
ts.
c P�
0.05
for
diff
eren
cein
rate
s.d Sa
mpl
esi
zen
otcl
ear.
e Con
stip
atio
nde
fin
edas
bow
elm
ovem
ent
less
freq
uen
tly
than
ever
ytw
oda
ys.
f Res
ults
from
end
offi
rst
wee
kof
trea
tmen
tbe
caus
eof
hig
hra
teof
wit
hdr
awal
saf
ter
firs
tw
eek.
g Res
ults
repo
rted
on10
cmvi
sual
anal
ogsc
ale.
h Dos
e-lim
itin
gsi
deef
fect
s(n
otw
ith
draw
alra
te),
P�
0.00
3fo
rdi
ffer
ence
inra
tes.
Vol. 26 No. 5 November 2003 1041Long-Acting Oral Opioids for Chronic Non-Cancer Pain
Tab
le5
Stud
yC
hara
cter
isti
csan
dA
dver
seE
vent
s,O
bser
vati
onal
Stud
ies
ofL
ong-
Act
ing
Opi
oids
Con
fusi
onor
Lon
g-A
ctin
gQ
ualit
yD
row
sin
ess
Dif
ficu
lty
Lon
g-T
erm
Stud
yO
pioi
dsSt
udie
dR
atin
gaN
ause
aV
omit
ing
Con
stip
atio
nor
Som
nol
ence
Diz
zin
ess
Con
cen
trat
ing
Wit
hdr
awal
bU
se
Ark
inst
all
1995
Lon
g-ac
tin
gPO
OR
(2)
NR
NR
NR
NR
NR
NR
NR
54%
(15/
28)
code
ine
Bac
h20
01L
ong-
acti
ng
POO
R(0
)N
RN
RN
RN
RN
RN
RN
RN
Rm
orph
ine
(tw
ice-
daily
)C
aldw
ell
2002
Lon
g-ac
tin
gPO
OR
(2)
16%
6%35
%13
%9%
NR
33%
48%
mor
phin
e(2
9/18
1)(1
1/18
1)(6
3/18
1)(2
3/18
1)(1
6/18
1)(6
0/18
1)(8
6/18
1)(o
nce
-dai
ly)
Del
lem
ijn19
98T
ran
sder
mal
POO
R(2
)92
%56
%36
%58
%54
%�
20%
NR
19%
fen
tan
yl(9
/48)
Dun
bar
1996
Met
had
one
POO
R(0
)N
RN
RN
RN
RN
RN
RN
RN
RL
ong-
acti
ng
NR
NR
NR
NR
NR
NR
NR
NR
mor
phin
eG
reen
1996
Met
had
one
POO
R(0
)N
RN
RN
RN
RN
RN
RN
RN
RM
illig
an20
01T
ran
sder
mal
POO
R(1
)9%
8%N
RN
RN
RN
RN
RN
Rfe
nta
nyl
(48/
530)
(42/
530)
Rot
h20
00L
ong-
acti
ng
FAIR
(4)
24%
NR
52%
30%
NR
NR
30%
43%
oxyc
odon
e(2
5/10
6)(5
5/10
6)(3
2/10
6)(3
2/10
6)(4
6/10
6)a N
umbe
rof
crit
eria
out
ofse
ven
adeq
uate
lym
et.
b Due
toad
vers
eev
ents
.N
R�
not
repo
rted
.
1042 Vol. 26 No. 5 November 2003Chou et al.
Tab
le6
Com
para
tive
Res
ults
for
Adv
erse
Eve
nts,
Tri
als
ofL
ong-
Act
ing
Opi
oid
vers
usSh
ort-A
ctin
gO
pioi
d
Dro
wsi
nes
sSt
udy
Nau
sea
Vom
itin
gC
onst
ipat
ion
orSo
mn
olen
ceD
izzi
nes
sC
onfu
sion
Wit
hdr
awal
a
Lon
g-ac
ting
oxyc
odon
eC
aldw
ell
1999
Favo
rslo
ng-
acti
ng
Favo
rslo
ng-
acti
ng
Favo
rssh
ort-a
ctin
gFa
vors
lon
g-ac
tin
gFa
vors
lon
g-ac
tin
gN
otre
port
edFa
vors
lon
g-ac
tin
gH
aleb
1999
Favo
rslo
ng-
acti
ng
No
diff
eren
ceFa
vors
lon
g-ac
tin
gFa
vors
shor
t-act
ing
Favo
rssh
ort-a
ctin
gN
otre
port
edN
odi
ffer
ence
Salz
man
b19
98Fa
vors
shor
t-act
ing
Favo
rssh
ort-a
ctin
gFa
vors
lon
g-ac
tin
gFa
vors
lon
g-ac
tin
gFa
vors
shor
t-act
ing
No
diff
eren
ceFa
vors
shor
t-act
ing
Oth
erlo
ng-a
ctin
gop
ioid
sG
osti
ck19
89N
otre
port
edN
otre
port
edFa
vors
lon
g-ac
tin
gN
otre
port
edN
otre
port
edN
otre
port
edN
otre
port
edH
alec
1997
Favo
rssh
ort-a
ctin
gFa
vors
shor
t-act
ing
No
diff
eren
ceFa
vors
shor
t-act
ing
Favo
rssh
ort-a
ctin
gN
otre
port
edFa
vors
shor
t-act
ing
Llo
yd19
92Fa
vors
shor
t-act
ing
Not
repo
rted
No
diff
eren
ceFa
vors
shor
t-act
ing
Not
repo
rted
Favo
rssh
ort-a
ctin
gFa
vors
shor
t-act
ing
Jam
ison
c19
98Fa
vors
shor
t-act
ing
Not
repo
rted
Favo
rssh
ort-a
ctin
gFa
vors
shor
t-act
ing
Favo
rslo
ng-
acti
ng
No
diff
eren
ceN
otre
port
eda D
ueto
adve
rse
even
t.b St
udie
dsa
me
popu
lati
on.
c Low
erdo
seof
opio
idus
edin
shor
t-act
ing
arm
.
comparisons of one long-acting opioid with an-other. All were rated fair quality for generalmethodology and poor or fair quality for adverseevent assessment (trial quality reviewed in previ-ous sectionsof this report).Subgroupsof trials forspecific types of pain have the same problemswith heterogeneity in interventions, outcomesassessed, and findings that were encountered inexamining general efficacy and adverse events.They are further limited by the smaller numberof available trials for each type of pain. Thesetrials provide insufficient indirect evidence thatone long-acting opioid is superior to any other inany subpopulation of patients with chronic pain.
It is not possible to draw reliable conclusionsregarding comparative efficacy or adverse eventrates for any subpopulation from these data.
Summary of ResultsResults for each of the key questions are sum-
marized in Table 7. It is important to note thatwe identified no trials investigating methadoneor levorphanol in adult patients with chronicnon-cancer pain. The results refer to studiesthat investigated transdermal fentanyl and long-acting oral oxycodone, morphine, codeine,and dihydrocodeine.
In general, there was insufficient evidenceto prove that different long-acting opioids areassociated with different efficacy or adverseevent rates. Only one poor-quality trial20 di-rectly compared different long-acting opioids(transdermal fentanyl and long-acting mor-phine) and gave inconclusive results. This wasthe only trial we identified that evaluatedtransdermal fentanyl in patients with non-cancer pain. This trial may show that transder-mal fentanyl is a reasonable second choicefor patients who have inadequate pain reliefon morphine, but does not answer the generalquestion of which long-acting opioid is superiorfor the general population of patients withchronic non-cancer pain. It also did not provideconvincing evidence that transdermal fentanylis associated with less constipation than oralmorphine, as has been consistently found intrials of cancer patients.42 Another fair-qualitytrial21 that directly compared once-dailyversus twice-daily morphine also gave inconclu-sive results. Although this study found a slightimprovement in overall quality of sleep for once-daily morphine given in the morning compared
Vol. 26 No. 5 November 2003 1043Long-Acting Oral Opioids for Chronic Non-Cancer Pain
Tab
le7
Sum
mar
yof
Evi
denc
e
Key
Que
stio
ns
Lev
elof
Evi
den
ceC
oncl
usio
ns
Effic
acy
1A.
Inh
ead-
to-h
ead
com
pari
son
s,h
ason
eor
mor
elo
ng-
POO
RM
ost
lon
g-ac
tin
gop
ioid
sh
ave
not
been
com
pare
ddi
rect
lyin
clin
ical
tria
ls.
Tw
otr
ials
acti
ng
opio
idbe
ensh
own
tobe
supe
rior
toot
her
lon
g-di
rect
lyco
mpa
red
one
lon
g-ac
tin
gop
ioid
toan
oth
er.
On
epo
or-q
ualit
yst
udy
(lac
kof
acti
ng
opio
ids
inre
duci
ng
pain
and
impr
ovin
gfu
nct
ion
albl
indi
ng,
hig
hpr
opor
tion
ofpa
tien
tson
stud
ydr
ugpr
ior
toen
try,
hig
hlo
ssto
follo
w-
outc
omes
wh
enus
edfo
rtr
eatm
ent
ofad
ults
wit
hch
ron
icup
)di
rect
lyco
mpa
red
one
lon
g-ac
tin
gop
ioid
(tra
nsd
erm
alfe
nta
nyl
)to
anot
her
non
-can
cer
pain
?(m
orph
ine)
.O
ne
fair
-qua
lity
stud
yco
mpa
red
diff
eren
tlo
ng-
acti
ng
form
ulat
ion
s(o
nce
-or
twic
e-da
ily)
ofm
orph
ine,
foun
dn
osi
gnif
ican
tdi
ffer
ence
inpa
inco
ntr
olan
da
sign
ific
ant
diff
eren
cefo
ron
eof
seve
nm
easu
res
ofsl
eep
qual
ity
usin
gon
ce-d
aily
mor
phin
ein
the
a.m
.,bu
tn
otp.
m.
Th
ere
isin
suff
icie
nt
evid
ence
from
hea
d-to
-hea
dco
mpa
riso
nst
udie
sto
sugg
est
that
one
lon
g-ac
tin
gop
ioid
issu
peri
orto
anot
her
inte
rms
ofef
fica
cyin
adul
tpa
tien
tsw
ith
chro
nic
non
-can
cer
pain
.1B
.In
tria
lsco
mpa
rin
glo
ng-
acti
ng
opio
ids
toot
her
type
sPO
OR
Four
teen
tria
lsco
mpa
relo
ng-
acti
ng
opio
ids
toot
her
type
sof
drug
sor
topl
aceb
o.T
hey
are
ofdr
ugs
orto
plac
ebo,
isth
ere
apa
tter
nto
sugg
est
that
too
het
erog
eneo
usan
dof
insu
ffic
ien
tly
hig
hqu
alit
yto
com
pare
the
effi
cacy
oflo
ng-
one
lon
g-ac
tin
gop
ioid
ism
ore
effe
ctiv
eth
anan
oth
er?
acti
ng
opio
ids.
Th
ere
isin
suff
icie
nt
evid
ence
tosu
gges
tth
aton
elo
ng-
acti
ng
opio
idis
supe
rior
toan
oth
erin
term
sof
effi
cacy
inad
ult
pati
ents
wit
hch
ron
icn
on-c
ance
rpa
in.
1C.
Hav
elo
ng-
acti
ng
opio
ids
been
show
nto
besu
peri
orto
POO
RSe
ven
fair
-qua
lity
tria
lsdi
rect
lyco
mpa
reth
eef
fica
cyof
lon
g-an
dsh
ort-a
ctin
gop
ioid
sin
shor
t-act
ing
opio
ids
inre
duci
ng
pain
and
impr
ovin
gpa
tien
tsw
ith
chro
nic
non
-can
cer
pain
.T
hes
etr
ials
wer
eh
igh
lyh
eter
ogen
eous
,in
term
sfu
nct
ion
alou
tcom
esw
hen
used
for
trea
tmen
tin
adul
tsw
ith
ofst
udy
desi
gn,
pati
ent
popu
lati
ons,
inte
rven
tion
s,an
dou
tcom
esas
sess
ed.
chro
nic
non
-can
cer
pain
?T
her
eis
insu
ffic
ien
tev
iden
ceto
sugg
est
supe
rior
effi
cacy
oflo
ng-
acti
ng
opio
ids
asa
clas
sco
mpa
red
tosh
ort-a
ctin
gop
ioid
sin
adul
tsw
ith
chro
nic
non
-can
cer
pain
.T
hre
eof
the
tria
lsco
mpa
relo
ng-
acti
ng
oxyc
odon
eto
shor
t-act
ing
oxyc
odon
ean
dw
ere
mor
eh
omog
eneo
us.
Non
efo
und
diff
eren
ces
incl
inic
alef
fica
cy.
Th
ere
isfa
irev
iden
ceto
sugg
est
that
lon
g-ac
tin
gox
ycod
one
and
shor
t-act
ing
oyxc
odon
ear
eeq
ually
effe
ctiv
efo
rpa
inco
ntr
olin
adul
tpa
tien
tsw
ith
chro
nic
non
-can
cer
pain
.A
dver
seEv
ents
2A.
Inh
ead-
to-h
ead
com
pari
son
s,h
ason
eor
mor
elo
ng-
POO
RM
ost
lon
g-ac
tin
gop
ioid
sh
ave
not
been
com
pare
ddi
rect
lyin
clin
ical
tria
ls.
On
epo
or-
acti
ng
opio
idbe
ensh
own
tobe
asso
ciat
edw
ith
few
erqu
alit
ytr
ial
(see
abov
e)di
rect
lyco
mpa
res
one
lon
g-ac
tin
gop
ioid
wit
han
oth
eran
don
ead
vers
eev
ents
com
pare
dto
oth
erlo
ng-
acti
ng
opio
ids
wh
enfa
ir-q
ualit
ytr
ial
(see
abov
e)di
rect
lyco
mpa
res
once
-dai
lyw
ith
twic
e-da
ilym
orph
ine.
Inus
edfo
rtr
eatm
ent
ofad
ults
wit
hch
ron
icn
on-c
ance
rpa
in?
the
fair
-qua
lity
tria
l,on
ce-d
aily
mor
phin
ew
asas
soci
ated
wit
ha
hig
her
freq
uen
cyof
con
stip
atio
nan
da
low
erfr
eque
ncy
ofas
then
iaco
mpa
red
totw
ice-
daily
mor
phin
e;ot
her
adve
rse
even
tra
tes
wer
en
otsi
gnif
ican
tly
diff
eren
t.T
her
eis
insu
ffic
ien
tev
iden
ceto
sugg
est
that
one
lon
g-ac
tin
gop
ioid
issu
peri
orin
term
sof
adve
rse
even
tsth
anan
yot
her
inad
ult
pati
ents
wit
hch
ron
icn
on-c
ance
rpa
in.
2B.
Intr
ials
com
pari
ng
lon
g-ac
tin
gop
ioid
sto
oth
erty
pes
POO
RT
hir
teen
tria
lsco
mpa
relo
ng-
acti
ng
opio
ids
toot
her
type
sof
drug
sor
plac
ebo.
Th
ese
tria
lsof
drug
sor
topl
aceb
o,is
ther
ea
patt
ern
tosu
gges
tth
atar
eto
oh
eter
ogen
eous
and
ofin
suff
icie
ntl
yh
igh
qual
ity
tode
term
ine
rela
tive
risk
ofon
elo
ng-
acti
ng
opio
idis
asso
ciat
edw
ith
few
erad
vers
eas
sess
edad
vers
eev
ents
.R
ates
ofab
use
and
addi
ctio
nw
ere
not
repo
rted
intr
ials
.ev
ents
than
anot
her
?O
bser
vati
onal
stud
ies
onad
vers
eev
ent
wer
eof
gen
eral
lypo
orer
qual
ity
than
the
clin
ical
tria
ls.
Th
ere
isin
suff
icie
nt
evid
ence
tosu
gges
tth
aton
elo
ng-
acti
ng
opio
idis
supe
rior
inte
rms
ofad
vers
eev
ents
than
any
oth
erin
adul
tpa
tien
tsw
ith
chro
nic
non
-can
cer
pain
.Su
bpop
ulat
ions
3.A
reth
ere
subp
opul
atio
ns
ofpa
tien
ts(s
peci
fica
llyra
ce,
POO
RT
her
eis
alm
ost
no
info
rmat
ion
rega
rdin
gth
eco
mpa
rati
veef
fica
cyof
lon
g-ac
tin
gop
ioid
sag
e,se
x,or
type
ofpa
in)
wit
hch
ron
icn
on-c
ance
rpa
info
rfo
rsp
ecif
icsu
bpop
ulat
ion
sas
char
acte
rize
dby
race
,ge
nde
r,or
age.
For
spec
ific
type
sof
wh
ich
one
lon
g-ac
tin
gop
ioid
ism
ore
effe
ctiv
eor
asso
ciat
edch
ron
icn
on-c
ance
rpa
in,
fin
din
gsar
elim
ited
bypr
oble
ms
wit
hin
tern
alva
lidit
y,ex
tern
alw
ith
few
erad
vers
eef
fect
s?va
lidit
y,h
eter
ogen
eity
,an
dsm
all
num
bers
oftr
ials
for
each
subp
opul
atio
n.
Itis
not
poss
ible
todr
awre
liabl
eco
ncl
usio
ns
rega
rdin
gco
mpa
rati
veef
fica
cyor
adve
rse
even
tra
tes
for
any
subp
opul
atio
nfr
omth
ese
data
.
1044 Vol. 26 No. 5 November 2003Chou et al.
to twice-daily morphine, it also found signifi-cantly more constipation in the once-daily mor-phine group (though less asthenia). Othermeasures of sleep quality and pain control werenot significantly different. Studies that providedindirect data were too heterogeneous in terms ofstudy design, patient populations, interventions,assessed outcomes, and results to make accuratejudgments regarding comparative efficacy or ad-verse event rates. The comparative efficacyand adverse event rates of different long-actingopioids in adult patients with chronic non-cancer pain remains uncertain. In general, insuf-ficient data was provided in the included trialsto accurately assess external validity.
There was also insufficient evidence from asubgroup of seven trials to determine whetherlong-acting opioids as a class are more effectiveor associated with fewer adverse events thanshort-acting opioids. Three trials investigatinglong-acting oxycodone versus short-acting oxy-codone23,25,28 were more homogeneous andprovided fair evidence that long-acting andshort-acting oxycodone are equally effective forpain control. It is not clear whether recentmedia attention and case reports of abuse, ad-diction, and overdose (including respiratory de-pression) from long-acting opioids represent atrue increased risk or are proportionate to pre-scribing pattern changes.1 There also may beother reasons (such as convenience, improvedcompliance, or more consistent pain relief) forprescribing long-acting opioids, but these out-comes were not assessed in the reviewed trials.
Essentially no good-quality data are avail-able to assess comparative efficacy and adverseevent risks in subpopulations of patients withchronic non-cancer pain.
DiscussionThe current available literature does not pro-
vide enough evidence to guide the prescribingphysician in choosing an initial long-actingopioid medication for patients with chronic non-cancer pain. The lack of high-quality evidencecomparing long-acting opioids to one anotherand to short-acting opioids in patients withchronic non-cancer pain is concerning given thewide use of this class of medication in this popu-lation. Data are inadequate to determinewhether long-acting opioid preparations, eithercompared to each other or to short-acting opi-oids, have different efficacy and safety profiles.
Given the ever-rising costs of medications, itis likely that pressure will continue to increaseto watch the bottom line. Without good qualityevidence of comparative efficacy and safety,payers may be compelled to rely on cost as theonly method of differentiating between medica-tions in this drug class, or extrapolate fromstudies performed in other populations (e.g.,cancer pain patients) that may not be applicableto the population in question.
Ideal studies to investigate comparative effi-cacy and safety would perform head-to-headcomparisons of equianalgesic doses of long-acting opioids to other long-acting or short-acting opioids, be adequately blinded, use a pre-defined and systematic method for identifyingadverse events, be of longer duration, and ac-count for prior opioid use of enrollees. Large,population-based observational studies wouldhelp determine whether rare but serious adverseevent rates (such as respiratory depression) dif-fered between long-acting opioids. Particular at-tention to the risks of abuse and addiction wouldalso be best obtained from high-quality cohortstudies, as trials have typically excluded patientsat high risk for these complications. Outcomesshould be standardized or measured using a vari-ety of visual analogue scales, categorical scales,and common multiquestion assessments, so thatresults can be meaningfully compared acrossstudies. Another area that deserves careful studyis the efficacy of opioid rotation in this group ofpatients. We hope this report helps to highlightremaining gaps in our understanding of this im-portant class of medication and that studies tofill these gaps will be supported and undertaken.
AcknowledgmentsThe authors wish to acknowledge the Oregon
Department of Human Resources for its fund-ing support. They also wish to acknowledge theadministrative support provided by KathrynPyle Krages, AMLS, MA, Susan Wingenfeld,Susan Carson, MPH, and Patty Davies, MS.Additional information regarding Oregon’sPractitioner-Managed Prescription Drug Planis available at http://www.ohpr.state.or.us.
References1. Joranson DE, Ryan KM, Gilson AM, et al. Trends
in medical use and abuse of opioid analgesics. JAMA2000;283(13):1710–1714.
Vol. 26 No. 5 November 2003 1045Long-Acting Oral Opioids for Chronic Non-Cancer Pain
2. Strumpf M, Dertwinkel R, Wiebalck A, et al. Roleof opioid analgesics in the treatment of chronic non-cancer pain. CNS Drugs 2000;14(2):147–155.
3. Anonymous. The use of opioids for the treat-ment of chronic pain. A consensus statement fromthe American Academy of Pain Medicine and theAmerican Pain Society. Clin J Pain 1997;13(1):6–8.
4. Anonymous. The management of chronic painin older persons. AGS Panel on Chronic Pain in OlderPersons. American Geriatrics Society. Geriatrics1998;53(Suppl 3):S8–24.
5. Schug SA, Merry AF, Acland RH. Treatmentprinciples for the use of opioids in pain of nonmalig-nant origin. Drugs 1991;42(2):228–239.
6. Portenoy RK. Opioid therapy for chronic non-malignant pain: a review of the critical issues. JPain Symptom Manage 1996;11(4):203–217.
7. McQuay HJ. Opioid use in chronic pain. Bando-lier 2002; available at http://www.jr2.ox.ac.uk/bandolier/booth/painpag/wisdom/S31.html
8. Rinaldi RC, Steindler EM, Wilford BB, et al. Clar-ification and standardization of substance abuseterminology. JAMA 1988;259:555–557.
9. Savage SR. Opioid use in the management ofchronic pain. Med Clin North Am 1999;83(3):761–786.
10. Roth SH, Fleischmann RM, Burch FX, et al.Around-the-clock, controlled-release oxycodonetherapy for osteoarthritis-related pain: placebo con-trolled trial and long-term evaluation. Arch InternMed 2000;160:853–860.
11. Anonymous. Undertaking systematic reviews ofresearch on effectiveness: CRD’s guidance for thosecarrying out or commissioning reviews CRD ReportNumber 4 (2nd edition). York, UK: NHS Centrefor Reviews and Dissemination; 2001. Report No. 4(2nd edition).
12. Harris RP, Helfand M, Woolf SH, et al. Currentmethods of the third U.S. Preventive Services TaskForce. Am J Prev Med 2000;20(3S):21–35.
13. Mulrow CD, Oxman A. How to conduct a Coch-rane systematic review. Version 3.0.2. In: San AntonioCochrane Collaboration; 1997.
14. Quang-Cantagrel ND, Wallace MS, MagnusonSK. Opioid substitution to improve the effectivenessof chronic noncancer pain control: a chart review.Anesth Analg 2000;90(4):933–937.
15. Thomsen AB, Becker N, Eriksen J. Opioid rota-tion in chronic nonmalignant pain patients: a retro-spective study. Acta Anaesthesiol Scand 1999;43:918–923.
16. Bruera E, Pereira J, Watanabe S, et al. Opioidrotation in patients with cancer pain. Cancer 1996;78(4):852–857.
17. de Stoutz ND, Bruera E, Suarez-Almazor M. Opioidrotation for toxicity reduction in terminal cancer pa-tients. J Pain Symptom Manage 1995;10(5):378–384.
18. Raja SN, Haythornthwaite JA, Pappagallo M, etal. Opioids versus antidepressants in postherpeticneuralgia: a randomized, placebo-controlled trial.Neurology 2002;59:1015–1021.
19. Levy MH. Pharmacologic treatment of cancerpain. N Engl J Med 1996;335:1124–1132.
20. Allan L, Hays H, Jensen NH, et al. Randomizedcrossover trial of transdermal fentanyl and sus-tained-release oral morphine for treating chronicnon-cancer pain. BMJ 2001;322:1154–1158.
21. Caldwell JR, Rapoport RJ, Davis JC, et al. Efficacyand safety of a once-daily morphine formulation inchronic, moderate-to-severe osteoarthritis pain: re-sults from a randomized, placebo-controlled,double-blind trial and an open-label extension trial.J Pain Symptom Manage 2002;23:279–291.
22. Hale M, Speight K, Harsanyi Z, et al. Efficacy of12 hourly controlled-release codeine compared withas required dosing of acetaminophen plus codeinein patients with chronic low back pain. Pain ResManage 1997;2:33–38.
23. Caldwell JR, Hale ME, Boyd RE, et al. Treatmentof osteoarthritis pain with controlled- release oxyco-done or fixed combination oxycodone plus acet-aminophen added to nonsteroidal anti-inflammatorydrugs: a double blind, randomized, multicenter, pla-cebo controlled trial. J Rheumatol 1999;26:862–869.
24. Gostick N, Allen J, Cranfield R, et al. A compari-son of the efficacy and adverse effects of control-led-release dihydrocodeine and immediate-releasedihydrocodeine in the treatment of pain in osteoar-thritis and chronic back pain. In: Twycross RG, ed.Proceedings of The Edinburgh Symposium on PainControl and Medical Education; 1989:137–143.
25. Hale ME, Fleischmann R, Salzman R, et al. Effi-cacy and safety of controlled-release versus imme-diate-release oxycodone: randomized, double blindevaluation in patients with chronic back pain. ClinJ Pain 1999;15:179–183.
26. Jamison RN, Raymond SA, SlawsbyEA, et al. Opioidtherapy for chronic noncancer back pain. A random-ized prospective study. Spine 1998;23:2591–2600.
27. Lloyd RS, Costello F, Eves MJ, et al. The efficacyand tolerability of controlled-release dihydrocodeinetablets and combination dextropropoxyphene/para-cetamol tablets in patients with severe osteoarthritisof the hips. Curr Med Res Opin 1992;13:37–48.
28. Salzman RT, Roberts MS, Wild J, et al. Can acontrolled release oral dose form of oxycodone beused as readily as an immediate-release form forthe purpose of titrating to stable pain control? J PainSymptom Manage 1999;18:271–279.
29. Arkinstall W, Sandler A, Goughnour B, et al.Efficacy of controlled-release codeine in chronicnonmalignant pain: a randomized, placebo con-trolled clinical trial. Pain 1995;62:169–178.
30. Harke H, Gretenkort P, Ladleif HU, et al. Theresponse of neuropathic pain and pain in complex
1046 Vol. 26 No. 5 November 2003Chou et al.
regional pain syndrome I to carbamazepine and sus-tained-release morphine in patients pretreated withspinal cord stimulation: a double-blinded random-ized study. Anesth Analg 2001;92:488–495.
31. Huse E, Larbig W, Flor H, et al. The effect ofopioids on phantom limb pain and cortical reorgani-zation. Pain 2001;90:47–55.
32. Moulin DE, Iezzi A, Amireh R, et al. Randomizedtrial of oral morphine for chronic noncancer pain.Lancet 1996;347:143–147.
33. Peloso PM, Bellamy N, Bensen W, et al. Doubleblind randomized placebo control trial of controlled-release codeine in the treatment of osteoarthritis ofthe hip or knee. J Rheumatol 2000;27(3):764–771.
34. Watson CP, Babul N. Efficacy of oxycodone inneuropathic pain a randomized trial in postherpeticneuralgia. Neurology 1998;50:1837–1841.
35. Moran C. MST continuous tablets and pain con-trol in severe rheumatoid arthritis. Br J Clin Res1991;2:1–12.
36. Dellemijn PL, van Duijn H, Vanneste JA. Pro-longed treatment with transdermal fentanyl in neu-ropathic pain. J Pain Symptom Manage 1998;16:220–229.
37. Dunbar SA, Katz NP. Chronic opioid therapyfor nonmalignant pain in patients with a history ofsubstance abuse: report of 20 cases. J Pain SymptomManage 1996;11:163–171.
38. Green J, Hickey S, Ansbacher S, et al. Methadoneuse in a small series of selected patients with chronicnonmalignant pain. Pain Digest 1996;6:3–6.
39. Milligan K, Lanteri-Minet M, Borchert K, et al.Evaluation of long-term efficacy and safety of trans-dermal fentanyl in the treatment of chronic noncan-cer pain. J Pain 2001;2:197–204.
40. Bach V, Kamp-Jensen M, Jensen NH, et al. Bu-prenorphine and sustained-release morphine: effectand side effects in chronic use. Pain Clin 1991;4:87–93.
41. Arkinstall WW, Goughnour BR, White JA, et al.Control of severe pain with sustained-release mor-phine tablets vs. oral morphine solution. Can MedAssoc J 1989;140:653–661.
42. Haazen L, Noorduin H, Megens A, et al. Theconstipation-inducing potential of morphine andtransdermal fentanyl. European J Pain 1999;3(Suppl.A):9–15.
Appendix ASearch Strategy
1 exp analgesics, opioid/or “opioid analgesics”.mp.2 exp narcotics/or “narcotics”.mp.3 1 or 24 (intractable pain or severe pain or chronic pain).mp.5 3 and 46 limit 5 to human7 limit 6 to english language8 6 not 79 limit 8 to abstracts
10 7 or 9
Appendix BMethods for Drug Class Reviews for Oregon
Health Plan Practitioner-Managed PrescriptionDrug Plan Oregon Health & Science University Evi-dence-Based Practice Center December 14, 2001
Available at http://www.oregonrx.org/OrgrxPDF/Opioid%20Review.htm or from the authors.
Vol. 26 No. 5 November 2003 1047Long-Acting Oral Opioids for Chronic Non-Cancer Pain
Appendix CQuality Abstraction Tool for Adverse Events of Opioids
Author Study
Year published
Citation
Setting (country, single or multicenter, specialty or primary care clinic)
Type of study (RCT, crossover, population-based, retrospective cohort, prospective cohort)
INTERNAL VALIDITY
Selection:1: Study states “all patients” or “consecutive series” during specified time period (observational study) or describes andaccounts for all patients deemed eligible (clinical trial) and has explicit inclusion and exclusion criteria applied to all eligiblepatients (all study types)0: Selection not clear, biased selection, inclusion and exclusion criteria not specified, or unable to determine proportion ofpatients eligible for trial who withdrew or were not entered
Loss to follow-up:1: Low overall and differential loss to follow-up (�15% of study population or �25% difference between groups), able tocompute adverse effects according to intention-to-treat if low loss to follow-up0: High overall or differential loss to follow-up (>15% overall or �25% difference between groups), or unable to calculateintention-to-treat if low loss to follow-up
Adverse events pre-specified and pre-defined:1: Study reports definitions used for assessed adverse events in an explicit, reproducible fashion0: Study does not meet above criteria
Ascertainment techniques adequately described:1: Study reports methods used to ascertain complications, including who ascertained, timing, and methods used0: Study does not meet above criteria
Non-biased and accurate ascertainment of adverse events:1: Patients and assessors blinded to intervention and ascertainment techniques go beyond patient self-report alone0: Study does not meet above criteria
Statistical analysis of potential confounders:1: Study examines more than 2 relevant confounders/risk factors using standard acceptable statistical techniques0: Study does not meet above criteria
Adequate duration of follow-up:1: Study reports duration of follow-up and duration at least 7 days0: Study does not meet above criteria
Internal validity score (0-7)
EXTERNAL VALIDITY
Adequate description of study population:1: Study reports 2 or more demographic characteristic and both basic clinical characteristics of pain syndrome and averageduration of pain0: Study does not meet above criteria
Does study report numbers screened and eligible (trial) or inception cohort (observational study)?
Are exclusion criteria specified and numbers excluded for each criteria reported?
Who is the funding source?
Are authors employed by the funding source?
Are data held by the funding source?
Are patients in the study on opioids prior to study entry?
1048 Vol. 26 No. 5 November 2003Chou et al.
Appendix DClinical Trials Search Results