Opioid therapy for treating rheumatoid arthritis pain (Review)

Transcription

1 Whittle SL, Richards BL, Husni E, Buchbinder R This is a reprint of a Cochrane review, prepared and maintained by The Cochrane Collaboration and published in The Cochrane Library 2011, Issue 11

2 T A B L E O F C O N T E N T S HEADER ABSTRACT PLAIN LANGUAGE SUMMARY SUMMARY OF FINDINGS FOR THE MAIN COMPARISON BACKGROUND OBJECTIVES METHODS RESULTS Figure Figure Figure DISCUSSION AUTHORS CONCLUSIONS ACKNOWLEDGEMENTS REFERENCES CHARACTERISTICS OF STUDIES DATA AND ANALYSES Analysis 1.1. Comparison 1 Opioid versus placebo, Outcome 1 Pain VAS Analysis 1.2. Comparison 1 Opioid versus placebo, Outcome 2 Mean daily pain intensity Analysis 1.3. Comparison 1 Opioid versus placebo, Outcome 3 Mean daily pain relief (maximum score = 4, negative score = pain worse) Analysis 1.4. Comparison 1 Opioid versus placebo, Outcome 4 Pain relief of 50% or better Analysis 1.5. Comparison 1 Opioid versus placebo, Outcome 5 Withdrawal due to inadequate analgesia Analysis 1.6. Comparison 1 Opioid versus placebo, Outcome 6 Patient reported global impression of clinical change (PGIC) good or very good Analysis 1.7. Comparison 1 Opioid versus placebo, Outcome 7 Desire to resume treatment Analysis 1.8. Comparison 1 Opioid versus placebo, Outcome 8 Functional status (HAQ) Analysis 1.9. Comparison 1 Opioid versus placebo, Outcome 9 Participants reporting adverse events Analysis Comparison 1 Opioid versus placebo, Outcome 10 Serious adverse events Analysis Comparison 1 Opioid versus placebo, Outcome 11 Withdrawal due to adverse events Analysis Comparison 1 Opioid versus placebo, Outcome 12 Withdrawal due to adverse event or inefficacy Analysis 2.1. Comparison 2 Opioid versus NSAID, Outcome 1 Change in 100mm pain VAS Analysis 2.2. Comparison 2 Opioid versus NSAID, Outcome 2 Global efficacy Analysis 2.3. Comparison 2 Opioid versus NSAID, Outcome 3 Desire to resume treatment Analysis 2.4. Comparison 2 Opioid versus NSAID, Outcome 4 Withdrawal due to inadequate analgesia Analysis 2.5. Comparison 2 Opioid versus NSAID, Outcome 5 Participants reporting adverse events Analysis 2.6. Comparison 2 Opioid versus NSAID, Outcome 6 Withdrawal due to adverse events APPENDICES WHAT S NEW HISTORY CONTRIBUTIONS OF AUTHORS DECLARATIONS OF INTEREST SOURCES OF SUPPORT DIFFERENCES BETWEEN PROTOCOL AND REVIEW INDEX TERMS i

3 [Intervention Review] Opioid therapy for treating rheumatoid arthritis pain Samuel L Whittle 1, Bethan L Richards 2, Elaine Husni 3, Rachelle Buchbinder 4 1 Rheumatology Unit, The Queen Elizabeth Hospital, Woodville, Australia. 2 Rheumatology, Royal Prince Alfred Hospital, Camperdown, Australia. 3 Department of Rheumatic and Immunologic Diseases, Cleveland Clinic: Orthopedic and Rheumatologic Institute, Cleveland, OH, USA. 4 Monash Department of Clinical Epidemiology at Cabrini Hospital, Department of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Malvern, Australia Contact address: Samuel L Whittle, Rheumatology Unit, The Queen Elizabeth Hospital, 28 Woodville Road, Woodville, South Australia, 5011, Australia. Samuel.Whittle@health.sa.gov.au. samuelwhittle@gmail.com. Editorial group: Cochrane Musculoskeletal Group. Publication status and date: New, published in Issue 11, Review content assessed as up-to-date: 3 May Citation: Whittle SL, Richards BL, Husni E, Buchbinder R. Opioid therapy for treating rheumatoid arthritis pain. Cochrane Database of Systematic Reviews 2011, Issue 11. Art. No.: CD DOI: / CD pub3. Background A B S T R A C T Despite improvements in the management of rheumatoid arthritis (RA), pain control is often inadequate even when inflammation is well controlled. Objectives To assess the efficacy and safety of opioid analgesics for treating pain in patients with RA. Search methods We searched the Cochrane Central Register of Controlled Trials (CENTRAL) (The Cochrane Library), MEDLINE and EMBASE for studies to May We also searched the 2008 to 2009 American College of Rheumatology (ACR) and European League against Rheumatism (EULAR) abstracts and performed a handsearch of the reference lists of articles. Selection criteria Studies were included if they were randomized or quasi-randomized controlled trials (RCTs or CCTs) which compared opioid therapy to another therapy (active or placebo) for pain in patients with RA. Outcomes of interest were pain, adverse effects, function and quality of life. Data collection and analysis Two review authors independently selected the studies for inclusion, extracted the data, and performed a risk of bias assessment. Main results Eleven studies (672 participants) were included in the review. Four studies assessed the efficacy of single doses of various opioid and non-opioid analgesics; a pooled analysis of these studies was not performed but in each study opioids reduced pain more than placebo. There were no differences between analgesic drugs in these studies. Seven studies were between one and six weeks in duration and assessed six different oral opioids (dextropropoxyphene, codeine, tramadol, tilidine, pentazocine, morphine), either alone or combined with non-opioid analgesics. The only strong opioid investigated 1

4 was controlled-release morphine sulphate, in a single study with 20 participants. Six studies compared an opioid to placebo. Opioids were superior to placebo in patient-reported global impression of change (3 studies, 324 participants: relative risk (RR) 1.44, 95% CI 1.03 to 2.03) but not for the number of withdrawals due to inadequate analgesia (4 studies, 345 participants: RR 0.82, 95% CI 0.34 to 2.0). Adverse events (most commonly nausea, vomiting, dizziness and constipation) were more frequent in patients receiving opioids compared to placebo (4 studies, 371 participants: odds ratio 3.90, 95% CI 2.31 to 6.56); the pooled risk ratio for withdrawal due to adverse events was 2.67 (3 studies, 331 participants: 95% CI 0.52 to 13.75). One study compared an opioid (codeine with paracetamol) to an NSAID (diclofenac) and found no difference in efficacy or safety between interventions. Authors conclusions There is limited evidence that weak oral opioids may be effective analgesics for some patients with RA, but adverse effects are common and may offset the benefits of this class of medications. There is insufficient evidence to draw conclusions regarding the use of weak opioids for longer than six weeks, or the role of strong opioids. P L A I N L A N G U A G E S U M M A R Y Opioid therapy for treating rheumatoid arthritis pain This summary of a Cochrane review presents what we know from research about the effect of opioids for treating rheumatoid arthritis pain. The review shows that in people with rheumatoid arthritis treated with weak opioids for up to six weeks: - Weak opioids may reduce pain compared with placebo - Treatment with weak opioids may result in more side-effects compared with placebo. There were no studies in people with rheumatoid arthritis that looked at the effects of weak opioids taken for more than six weeks. There were not enough studies of strong opioids to draw conclusions about their effects in rheumatoid arthritis. What is rheumatoid arthritis and what are opioids? When you have rheumatoid arthritis your immune system, which normally fights infection, attacks the lining of your joints. This makes your joints swollen, stiff and painful. The small joints of your hands and feet are usually affected first. There is no cure for rheumatoid arthritis at present, so the treatments aim to relieve the pain and stiffness and improve your ability to move. Opioids are powerful pain-relieving substances that range in strength from relatively mild, such as codeine, to strong, such as morphine. Some examples of weak opioids are codeine (for example Panadeine Forte ) and tramadol (for example Tramal). Some examples of strong opioids are oxycodone (for example Percocet, Endone), morphine and fentanyl (for example Duragesic). They can be taken in a pill form, as an injection or as a patch placed on the skin. Common opioid side-effects include nausea, constipation and drowsiness. Best estimate of what happens to people with rheumatoid arthritis who take opioids Patient-reported global impression of change -18 more people out of 100 reported a good or very good improvement in the symptoms of their rheumatoid arthritis after treatment with opioids for between one and six weeks (18% absolute improvement) -57 people out of 100 reported a good or very good improvement in symptoms -40 people out of 100 who took a placebo reported a good or very good improvement in symptoms Side-effects 2

5 S U M M A R Y O F F I N D I N G S F O R T H E M A I N C O M P A R I S O N [Explanation] Opioids for rheumatoid arthritis pain Patient or population: patients with rheumatoid arthritis pain Settings: Intervention:Opioids 1 Outcomes Illustrative comparative risks*(95% CI) Relative effect (95% CI) No of Participants (studies) Assumed risk Corresponding risk Control Opioids Patient reported global impression of clinical change (PGIC) good or very good PGIC scale: Very poor, Poor, No change, Good, Very good. Follow-up: 1-6 weeks 398per per1000 (410to808) RR 1.44 (1.03 to 2.03) 324 (3 studies) Withdrawal due to inadequate analgesia Follow-up: median 1.5 weeks 78per per1000 (27to157) RR 0.82 (0.34 to 2.01) 345 (4 studies) Functional status(haq) HAQ.Scalefrom:0to3. Follow-up: 1-2 weeks The mean Functional status(haq) in the intervention groups was 0.1 lower (0.33 lower to 0.13 higher) 243 (2 studies) Quality of the evidence (GRADE) Comments low 2,3 Absolute risk difference =18%(1%to41%).Relative percent change = 44% (3% to 103%). NNT =6(3to84) 4 Not statistically significant low 2 4 HAQonlyreportedintwo low 5 studies (of duration one week and two weeks) 3

6 Not statistically significant low 3, (3 studies) RR 2.67 (0.52 to 13.75) 53per per1000 (28to729) Withdrawal due to adverse events Follow-up: 1-6 weeks low 3,7 Absolute risk difference = 30% (17% to 42%). Relative percent change =143%(81%to203%). NNTH=4(3to6) (4studies 8 ) OR 3.90 (2.31 to 6.56) 209per per1000 (379to634) Participants reporting adverse events Follow-up: 1-6 weeks low 9,10 Two studies reported serious adverse events. In each study, one participant in the opioid group experienced an SAE. There were no deaths See comment See comment Not estimable 317 (2 studies) Serious adverse events Follow-up: 1 weeks *The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention(and its 95% CI). CI: Confidence interval; RR: Risk ratio; OR: Odds ratio; GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Lowquality:Furtherresearchisverylikelytohaveanimportantimpactonourconfidenceintheestimateofeffectandislikelytochangetheestimate. Very low quality: We are very uncertain about the estimate. 1 Therewasheterogeneityinthechoiceofopioidandtheuseofanadjunctivenon-opioidanalgesic.10ofthe11includedstudiesused a weak opioid. Oral morphine was the only strong opioid used. 2 Includedstudieswerepotentiallybiasedduetounclearorinadequatesequencegeneration,allocation concealment andmanagement of incomplete data. 3 Substantial secular changes in the management of rheumatoid arthritis may limit the applicability of the findings to patients in the current era. 4 Note:Numberneededtotreat(NNT)=n/awhenresultisnotstatisticallysignificant.NNTfordichotomousoutcomescalculatedusing Cates NNT calculator( 5 Onlytwostudiesreportedthisoutcome;Moran1991wasathighriskofbiasduetoachangeindesignresultingfromanunexpectedly 4highnumberofwithdrawals.Lee2006reportedHAQscoresforonly223ofthe267participants;noexplanationwasprovided.

7 6 Adequate data for this outcome was only available for three trials, none of which clearly described the randomization or allocation concealment methods. 7 Data for this outcome was available from four trials, none of which clearly described the randomization or allocation concealment methods. 8 Oneofthefourtrialswascross-overdesign 9 Bedi1969athighriskofbiasduetoinadequaterandomization,potentialforcarry-overeffects, andincompleteoutcomedata.single SAE(upper GI bleed) occurred in the combination treatment group in phase I. 10 OnlytwoSAEswerereportedinthe11includedstudies.Onlyonestudy(Lee2006)pre-specifiedSAEsasanoutcomemeasure.9of the 11 studies reported adverse events. xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx 5

8 B A C K G R O U N D Rheumatoid arthritis (RA) is a chronic, inflammatory, destructive joint disease that affects about 1% of the population (Alarcon 1995). RA can cause progressive joint destruction and deformity despite state-of-the-art treatment. The mainstays of therapy in RA, disease-modifying anti-rheumatic drugs (DMARDs) and non-steroidal anti-inflammatory drugs (NSAIDs), primarily target inflammatory pathways, however there are multiple contributors to the genesis of pain in RA in addition to inflammation, including joint damage or destruction and peripheral and central sensitization (McDougall 2006), which may not be responsive to treatment with these medications. Indeed, despite recent improvements in the management of RA, including earlier intervention, more aggressive use of DMARDs and the advent of the biologic disease-modifying therapies (bdmards), many patients with RA continue to experience musculoskeletal pain even when inflammation is well-controlled (Kvien 2004). Patients with RA perceive pain to be their predominant impairment (Minnock 2003) and report pain management as their highest priority (Heiberg 2002). Although opioids have been claimed to successfully control pain in RA, the potential for adverse effects, addiction and drug interactions has limited their regular use (Fitzcharles 2009). Opioids produce their pharmacological effects via interaction with receptors located on neuronal cell walls. The synthetic analgesic tramadol combines an opioid receptor effect with monoamine reuptake inhibition (Raffa 1992). Opioids are commonly classified as either weak or strong, however no consensus exists for such a classification and there is no clear pharmacological distinction at the receptor level between drugs to which either of these labels are commonly applied. Opioids classified as weak tend to have maximal doses beyond which further escalation results in no increase in analgesic effect, or the onset of intolerable adverse effects, and are often administered as fixed dose combinations with nonopioid analgesics such as paracetamol (Goldman 2006). Because the analgesic properties of opioids do not depend on the presence of active inflammation, they are theoretically attractive as an analgesic option in RA patients with persistent pain despite the use of DMARDs. The presence of opioid receptors on peripheral afferent nerves within joints suggests that opioids may have peripheral analgesic effects in RA in addition to their effect on the CNS (Lang 2010). In addition, intra-articular nociceptive nerve fibres may contribute to joint inflammation via the release of neuropeptides such as substance P and calcitonin gene-related peptide, which raises the possibility that opioids may also act to inhibit joint inflammation (Fields 1980; McDougall 2006). The role of opioids in the management of acute pain, post-surgical pain and chronic cancer pain is clearly established (Hudcova 2006; Wiffen 2007). However, the role of opioid therapy for chronic non-cancer pain remains an area of debate. There is weak evidence that opioids may provide clinically significant pain relief in a proportion of individuals with chronic non-cancer pain, however many patients discontinue treatment due to adverse effects (Noble 2010). Although opioids are frequently used for the management of pain in RA sufferers (Grijalva 2008), current RA treatment guidelines offer scant guidance regarding the use of opioids in an evidence-based management algorithm (ACR 2002; Luqmani 2009). There is a large unmet need for safe and effective analgesic medications in this population, and this systematic review seeks to examine the evidence regarding the role of opioids in the management of RA pain. O B J E C T I V E S To assess the efficacy and safety of opioid analgesics for treating pain in patients with RA. M E T H O D S Criteria for considering studies for this review Types of studies All published randomized or quasi-randomized controlled trials (RCTs or CCTs) which compared opioid therapy to another therapy (placebo or active including non-pharmacological therapies) for RA were considered for inclusion. Studies of opioid therapy in RA in the immediate post-operative setting (for example after arthroplasty or arthroscopy) and studies that did not contain pain as an outcome measure were excluded. Only trials published as full articles or available as a full trial report were considered for inclusion. Types of participants Adult patients (aged 18 years or older) with a diagnosis of RA. Populations that included a mix of people with RA and other musculoskeletal pain were excluded unless results for the RA population could be separated out from the analysis. Types of interventions All trials that evaluated opioid medications were included. This includes tramadol and compound medications where an opioid is combined with a simple analgesic (such as codeine combined with paracetamol). Opioid administration via any route (including, but not limited to, oral, transdermal and intra-articular), any formulation (for example immediate-acting or sustained-release), and any interval (for example on-demand or at regular intervals) was included. Despite the lack of consensus regarding the boundary 6

9 between weak and strong opioids, the persistence of these descriptors in clinical practice suggests that clinicians and regulatory bodies find such a distinction to be a useful conceptual tool and we elected to retain such a classification. For the purpose of this review, we considered weak opioids to include codeine, tramadol, (dextro)propoxyphene, pentazocine and tilidine. Strong opioids included morphine, oxycodone, methadone, fentanyl, hydromorphone, buprenorphine and diamorphine. Comparators could be: 1. placebo; 2. a different opioid analgesic; 3. non-opioid analgesic medications (including NSAIDs, paracetamol, neuromodulators and antidepressants, but excluding DMARDs and bdmards); 4. non-pharmacological analgesic modalities (e.g. acupuncture, massage, transcutaneous electrical nerve stimulation (TENS)); 5. The same opioid medication via a different route, formulation or dosing interval. Types of outcome measures There is considerable variation in the outcome measures reported in clinical trials of interventions for pain. For the purpose of this systematic review, we aimed to include outcome measures that are considered to be of the greatest importance to patients with persistent pain and the clinicians who care for them. The Initiative on Methods, Measurement, and Pain Assessment in Clinical Trials (IMMPACT) has published consensus recommendations for determining clinically important changes in outcome measures in clinical trials of interventions for chronic pain. Reductions in pain intensity of 30% and 50% reflect moderate and substantial clinically important differences, respectively, and it is recommended that the proportion of patients that respond with these degrees of pain relief be reported (Dworkin 2008). Continuous outcome measures in pain trials (such as mean change on a 100 mm visual analogue scale) may not follow a Gaussian distribution. Often, a bimodal distribution is seen instead, where patients tend to report either very good or very poor pain relief (Moore 2010a). This creates difficulty in interpreting the meaning of average changes in continuous pain measures. For this reason, a dichotomous outcome measure (the proportion of participants reporting 30% pain relief) is likely to be more clinically relevant and was the primary efficacy measure in this review. It is recognised, however, that it has been the practice in most trials of interventions for chronic pain to report continuous measures and therefore the mean change in pain score was also included as a secondary efficacy measure. The pain state at the end of a clinical trial of an analgesic intervention, in contrast to measures of pain improvement, has also been recommended as a clinically relevant dichotomous outcome measure and was included as a secondary efficacy measure in this review (Moore 2010a). A global rating of treatment satisfaction, such as the Patient Global Impression of Change scale (PGIC) which provides an outcome measure that integrates pain relief, changes in function and side-effects into a single, interpretable measure, is also recommended by IMMPACT, and was included as a secondary outcome measure (Dworkin 2008). Primary outcomes 1. Efficacy: patient reported pain relief of 30% or greater. 2. Safety: number of withdrawals due to adverse events. Secondary outcomes 3. Pain: a. patient reported pain relief of 50% or greater; b. patient reported global impression of clinical change (PGIC) much or very much improved; c. proportion of patients achieving pain score below 30/100 mm on a visual analogue scale (VAS); d. mean change in pain score on a visual analogue scale or numerical rating scale. 4. Number and type of adverse events (AEs) and serious adverse events (SAEs, defined as AEs that are fatal, life-threatening, or require hospitalization). 5. Function, as measured by the Health Assessment Questionnaire (HAQ) or modified HAQ (Fries 1980; Pincus 1983). 6. Quality of life, as measured by either generic instruments (such as the Short Form-36 (SF-36)) or disease-specific tools (such as the Rheumatoid Arthritis Quality of Life instrument (RAQoL)). 7. Participant withdrawals due to inadequate analgesia. The duration of trials of interventions for pain varies considerably. The efficacy of interventions, and the relative balance of benefits and harms, may vary according to the duration of the trial and therefore the combination of results from trials of different durations may represent a source of bias in systematic reviews (Moore 2010a). For the purpose of this review, trials were grouped into those of duration < one week, one to six weeks, and > six weeks. A summary of findings table was generated that included the following outcomes of importance to patients: the proportion of patients reporting a PGIC of good or very good, the number of withdrawals due to inadequate analgesia, function, the number of participants reporting adverse events, total number of withdrawals due to adverse effects, and number of serious adverse events. Search methods for identification of studies Electronic searches We searched the following databases for RCTs or CCTs using the search strategies detailed in the appendices. 1. Ovid MEDLINE (1950 to May 2010) (Appendix 1). 2. EMBASE (1980 to May 2010) (Appendix 2). 7

10 3. Cochrane Central Register of Controlled Trials (CENTRAL) in The Cochrane Library (Appendix 3). No language restrictions were applied. Searching other resources Abstracts from the two major international rheumatology scientific meetings, the American College of Rheumatology (ACR) and the European League Against Rheumatism (EULAR), were also searched for the years 2008 and The reference lists of included articles were inspected for additional trials. Data collection and analysis Selection of studies All identified studies were assessed independently by two review authors (SW, BR) to identify the trials that fulfilled the inclusion criteria. All possibly relevant articles were retrieved in full text and any disagreement in study selection was resolved by consensus or by discussion with a third review author (RB or EH), as needed. Studies were translated into English where necessary. Data extraction and management Two review authors (SW and BR) independently extracted relevant information from the included trials including study design, characteristics of the study population, treatment regimen and duration, outcomes and timing of outcome assessment using predetermined forms. The raw data (means and standard deviations for continuous outcomes and number of events or participants for dichotomous outcomes) were extracted for the outcomes of interest. Differences in the data extracted were resolved by referring back to the original articles and establishing consensus. A third review author (RB or EH) was consulted to help resolve differences, if necessary. Assessment of risk of bias in included studies The potential for bias in the included studies was assessed using a risk of bias table. Two authors (SW, BR) independently assessed risk of bias for all included studies for the following items: random sequence generation; allocation concealment; blinding of participants, care provider, and outcome assessor for each outcome measure (see primary and secondary outcome measures); incomplete outcome data and other biases. This conformed to the methods recommended by The Cochrane Collaboration (Higgins 2009a). To determine the risk of bias of a study, for each criterion the presence of sufficient information and the likelihood of potential bias was evaluated. Each criterion was rated as Yes (low risk of bias), No (high risk of bias) or Unclear (either lack of information or uncertainty over the potential for bias). In a consensus meeting disagreements among the authors were discussed and resolved. A third review author (RB or EH) was available to make the final decision if no consensus had been reached but was not required. Measures of treatment effect The data were summarized in a meta-analysis only if there was sufficient clinical and statistical homogeneity. For continuous data, results were analysed as mean differences between the intervention and comparator group (MD), with corresponding 95% confidence intervals (CI). The mean difference between the treated group and the control group was weighted by the inverse of the variance in the pooled treatment estimate. For dichotomous data, a relative risk (RR) with corresponding 95% CI was calculated. Where dichotomous data from cross-over trials were combined with data from parallel-group trials using the generic inverse variance methods, the odds ratio (OR) with 95% CI was calculated rather than the RR. For studies containing more than two intervention groups, when making multiple pair-wise comparisons between all possible pairs of intervention groups possible we planned to include the same group of participants only once in the meta-analysis. Unit of analysis issues In the event that cross-over trials were identified in which the reporting of continuous outcome data precluded paired analysis, these data were not included in a meta-analysis in order to avoid unit-of-analysis error. Where carry-over effects were thought to exist, and where sufficient data existed, data from the first period only were included in the analysis (Higgins 2009b). Where dichotomous data from cross-over trials could be adequately extracted, a paired analysis (McNemar s test) was used to estimate the treatment effect (OR) and the corresponding standard error. Where appropriate, ORs from cross-over trials may be included in a meta-analysis with data from parallel-group trials using the generic inverse variance method; in all such cases a sensitivity analysis was performed that excluded data from cross-over trials (Elbourne 2002). McNemar s test was performed using Stata v9.2. Dealing with missing data Where important data were missing or incomplete, we planned to seek further information from the study authors, however this was not required for this review. In cases where individuals were missing from the reported results, we assumed the missing values to have a poor outcome. For dichotomous outcomes that measured adverse events (for example number of withdrawals due to adverse events), the withdrawal rate was calculated using the number of patients that received treatment as the denominator (worst case analysis). For dichotomous outcomes that measured benefits (for example proportion of participants achieving an ACR20 response), we used the worst case 8

11 analysis by using the number of randomized participants as the denominator. For continuous outcomes (for example pain), we calculated the MD or standardised mean difference (SMD) based on the number of patients analysed at the time point. If the number of patients analysed was not presented for each time point, the number of randomized patients in each group at baseline was to be used. Where possible, missing standard deviations were computed from other statistics such as standard errors, CIs or P values according to the methods recommended in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2009c). If standard deviations could not be calculated, they were imputed (for example from other studies in the meta-analysis) (Higgins 2009b). Assessment of heterogeneity Prior to meta-analysis, we assessed studies for clinical homogeneity with respect to type of therapy, control group, and the outcomes. For any studies judged as clinically homogeneous, statistical heterogeneity was assessed using the I 2 statistic (Deeks 2009) and using the following as a rough guide for interpretation: 0% to 40% might not be important heterogeneity, 30% to 60% may represent moderate heterogeneity, 50% to 90% may represent substantial heterogeneity and 75% to 100% considerable heterogeneity. In cases of considerable heterogeneity (defined as I 2 75%), we planned to explore the data further, including subgroup analyses, in an attempt to explain the heterogeneity; however statistical heterogeneity of this magnitude was not found in any of our metaanalyses. Assessment of reporting biases In order to determine whether reporting bias was present, we determined whether the protocol of the RCT was published before recruitment of the patients was started. For studies published after July 1st 2005, we screened the Clinical Trial Register at the International Clinical Trials Registry Platform of the World Health Organization ( (DeAngelis 2004). We evaluated whether selective reporting of outcomes was present (outcome reporting bias). We compared the fixed-effect model estimate against the randomeffects model to assess the possible presence of small sample bias in the published literature (that is where the intervention effect is more beneficial in smaller studies). In the presence of small sample bias, the random-effects model estimate of the intervention is more beneficial than the fixed-effect model estimate (Sterne 2008). The potential for reporting bias was to be further explored by funnel plots if 10 studies were available, however the heterogeneity of outcomes in the included studies precluded this form of analysis. Data synthesis Where studies were sufficiently homogeneous that it remained clinically meaningful for them to be pooled, meta-analysis was performed using a random-effects model regardless of the I 2 statistic results. Analysis was performed using Review Manager 5 and forest plots were produced for all analyses. Subgroup analysis and investigation of heterogeneity Where sufficient data were available, the following subgroup analyses were planned: 1. patients age (< 65 years versus 65 years); 2. opioid formulation (short-acting versus sustained-release); 3. route of opioid administration (oral versus parenteral); 4. duration of RA ( 2 years versus > 2 years). Sensitivity analysis Where sufficient studies existed, sensitivity analyses were planned to assess the impact of any bias attributable to inadequate or unclear treatment allocation (including studies with quasi-randomized designs). Presentation of key results A summary of findings table was produced using GRADEpro software. This table provides key information concerning the quality of evidence, the magnitude of effect of the interventions examined, and the sum of available data on the outcomes (short and long-term outcomes for pain and function, the total number of adverse events, serious adverse events and withdrawals due to adverse effects) as recommended by The Cochrane Collaboration (Schünemann 2008a). It includes an overall grading of the evidence related to each of the main outcomes using the GRADE approach (Schünemann 2008b). In addition to the absolute and relative magnitude of effect provided in the summary of findings table, for dichotomous outcomes the number needed to treat to benefit (NNTB) or the number needed to treat to harm (NNTH) was calculated, where appropriate. from the control group event rate (unless the population event rate was known) and the relative risk (RR) using the Visual Rx NNT calculator (Cates 2004). R E S U L T S Description of studies See: Characteristics of included studies; Characteristics of excluded studies; Characteristics of studies awaiting classification. See Characteristics of included studies, Characteristics of excluded studies. 9

12 Thirty-three possibly eligible studies were identified. Twenty-two studies were excluded; the reasons for exclusion are listed in the table Characteristics of excluded studies. Eleven studies (672 participants), published between 1969 and 2006, were included in the review (see flow chart in Figure 1). A full description of the included studies is provided in the table Characteristics of included studies. Ten of the included studies were published in English and one (Brunnmuller 2004) was published in German. Two further non-english studies (Tadashi 1972 in Japanese, Pavelka 1984 in Czech) were potentially eligible for inclusion based on their titles but were awaiting translation at the time of publication of this review. The review will be updated to include data from these studies if they are found to meet inclusion criteria. 10

13 Figure 1. Study flow diagram. 11

14 Nine studies were published prior to One study was conducted in the USA (Hardin 1979), one in Korea (Lee 2006) and the remainder in the United Kingdom or Europe. One study (Lee 2006) included 277 participants; the number of participants in the other 10 studies ranged from 16 to 78. Seven of the included studies used a cross-over design (Bedi 1969; Hardin 1979; Hill 1970; Huskisson 1974; Moran 1991; Nuki 1973; Saarialho 1988) and the other four studies used a parallel-group design (Boureau 1991; Brunnmuller 2004; Glowinski 1999; Lee 2006). In one of the cross-over studies (Moran 1991) only four of the 20 participants completed both phases of the trial so efficacy data were reported for the first phase only and this could be considered as a parallel-design trial for the purpose of analysis. All included studies recruited adult participants. Five studies (Boureau 1991; Brunnmuller 2004; Glowinski 1999; Lee 2006; Nuki 1973) reported the mean age of participants, which ranged from 52 to 58 years. The age range of participants was 21 to 79 years in the four studies (Brunnmuller 2004; Lee 2006; Nuki 1973; Saarialho 1988) where the range was reported. Four studies (Boureau 1991; Glowinski 1999; Nuki 1973; Saarialho 1988) reported the mean duration of RA in participants, which ranged from 6.2 to 8.8 years. In general, the degree of RA disease activity of participants was not reported. Only Brunnmuller 2004 reported the mean C-reactive protein (CRP) measurement at baseline (17mg/L). The proportion of participants receiving background NSAID therapy was not consistently described but, where reported, ranged from 63% (Brunnmuller 2004) to 100% (Nuki 1973; Saarialho 1988) of participants. Stable background therapy with NSAIDs was permitted to continue in all included studies except for three in which one of the study drugs was an NSAID (Hardin 1979; Hill 1970; Glowinski 1999). The use of background DMARDs by participants reflected the secular trends in RA management during this period. The use of DMARDs was excluded in only one study, published in 1973 (Nuki 1973). Stable DMARD therapy was permitted to continue in six studies (Hardin 1979; Boureau 1991; Brunnmuller 2004; Glowinski 1999; Lee 2006; Saarialho 1988). The proportion of participants using DMARDs was reported in only two studies: 88% (Glowinski 1999) and 94% (Lee 2006). In the remaining four studies (Bedi 1969; Hill 1970; Huskisson 1974; Moran 1991) DMARD use was not reported. None of the participants in the included studies were receiving biologic anti-rheumatic therapy. Three of the included studies were shorter than a week in duration (Hardin 1979; Hill 1970; Huskisson 1974). All were designed as cross-over trials with sequential administration of multiple interventions, including placebo and non-opioid analgesic drugs in addition to opioids. Each participant received one dose of each intervention in a randomized sequence. A fourth study (Saarialho 1988) also studied multiple analgesics in sequence. This study was designed to investigate psychomotor responses to analgesic agents in RA but also recorded pain as a secondary outcome. Each of the following four agents were administered to each participant for three days at two-week intervals: indomethacin, dextropropoxyphene, dextropropoxyphene plus amitriptyline, and placebo. Data from these four studies were not included in the meta-analysis. There were seven studies in which the study drug was administered for at least one week (Bedi 1969; Boureau 1991; Brunnmuller 2004; Glowinski 1999; Lee 2006; Moran 1991; Nuki 1973) but there were no studies identified with a follow-up period of longer than six weeks. Six of the seven studies (Bedi 1969; Boureau 1991; Brunnmuller 2004; Lee 2006; Moran 1991; Nuki 1973) were placebo-controlled and one (Glowinski 1999) compared a codeine-paracetamol compound plus low-dose diclofenac with standard-dose diclofenac alone. Four of the seven studies (Boureau 1991; Brunnmuller 2004; Glowinski 1999; Lee 2006) used a parallel-group design; Nuki 1973 and Bedi 1969 were cross-over trials and Moran 1991 was designed as a cross-over trial but reported the results of the first phase as a parallel-group trial due to an unexpectedly high rate of participant withdrawals. Six different opioids were investigated in the seven studies, most in combination with another non-opioid analgesic: a codeine and paracetamol combination was studied in both Boureau 1991 and Glowinski 1999; and the other five studies (Bedi 1969; Brunnmuller 2004; Lee 2006; Moran 1991; Nuki 1973) investigated tilidine with naloxone, tramadol with paracetamol, pentazocine, morphine sulphate and dextropropoxyphene respectively. Morphine (Moran 1991) was the only strong opioid investigated in the 11 included studies. Risk of bias in included studies The results of the risk of bias assessment are presented in Figure 2. 12

15 Figure 2. Risk of bias summary: review authors judgements about each risk of bias item for each included study. 13

16 Sequence generation and allocation concealment were insufficiently described for a clear judgement regarding bias in the majority of studies, and there was a high risk of biased randomization in at least two studies. Only Hardin 1979 clearly described an adequate method of allocation concealment. While most studies used a matching placebo, and all 11 studies were described as doubleblind, it was often unclear whether all of the study personnel were blinded to treatment allocation. Three of the studies that were at least one week in duration excluded drop-outs from the data analysis (Bedi 1969; Moran 1991; Nuki 1973). Of the four sequential single-dose studies, one (Huskisson 1974) presented data for fewer participants than had been randomized, but no description of drop-outs was provided; and in two (Hardin 1979, Hill 1970) data were presented only for those who completed the study and the number of drop-outs was not provided. There was a high risk of outcome reporting bias in five of the included studies (Boureau 1991; Brunnmuller 2004; Glowinski 1999; Hill 1970; Lee 2006). Effects of interventions See: Summary of findings for the main comparison Opioids for rheumatoid arthritis pain Due to differences in study duration and design, the seven studies that were at least one week in duration (Bedi 1969; Boureau 1991; Brunnmuller 2004; Glowinski 1999; Lee 2006; Moran 1991; Nuki 1973) were considered separately from the four trials that evaluated single doses of different analgesic drugs in sequence ( Hardin 1979; Hill 1970; Huskisson 1974; Saarialho 1988). Due to heterogeneity in study design and outcome measures in the latter four trials, a pooled analysis was not performed for these studies, which are instead described below. Participants in Hardin 1979 received analgesic-dose aspirin (650 mg), paracetamol, codeine sulphate, propoxyphene, pentazocine and placebo on consecutive days.the authors reported that single doses of all of the active agents resulted in superior pain relief compared to placebo. There were no significant differences in total pain relief between the active agents, although a significantly smaller proportion of those treated with propoxyphene achieved at least 50% reduction in pain compared with those treated with aspirin. Hill 1970 compared single doses of a compound of codeine phosphate, phenacetin and acetylsalicylic acid; two different doses of a compound of caffeine, phenacetin and isopropylantipyrine; and placebo administered sequentially on consecutive days. Pain relief was similar among the active treatments and all were superior to placebo. The study by Huskisson 1974 comprised three separate trials that each evaluated single doses of several analgesic drugs plus placebo. The study drugs included the opioids pentazocine, codeine (in combination with aspirin), dextropropoxyphene (in combination with paracetamol) and an experimental agent structurally related to propoxyphene (Ciba 44,328). Non-opioid study drugs included paracetamol and aspirin. All of the active treatments were associated with superior pain relief compared with placebo, but there was no consistent difference in efficacy between the active drugs. The primary outcome in the study by Saarialho 1988 was psychomotor performance in RA patients after treatment with analgesic medications, but pain scores were also recorded. Dextropropoxyphene at a dose of 130 mg had a greater effect on pain than either placebo or dextropropoxyphene 65 mg combined with amitriptyline 25 mg. Opioid versus placebo Six studies, with a total of 448 participants, had a duration of at least one week and investigated the effectiveness of opioids versus placebo (Bedi 1969; Boureau 1991; Brunnmuller 2004; Lee 2006; Moran 1991; Nuki 1973). The longest study (Brunnmuller 2004) was six weeks in duration. One study (Bedi 1969) compared the combination of dextropropoxyphene and aspirin with aspirin alone; the remainder compared an opioid with a matching placebo. In all six studies, participants who had been taking stable doses of NSAIDs were permitted to continue the background drug during the trial. Each of the six studies evaluated a different opioid and only one, morphine sulphate in Moran 1991, was considered as a strong opioid. The active intervention in three of the six studies was an opioid in combination with a non-opioid analgesic: codeine plus paracetamol (Boureau 1991), tramadol plus paracetamol (Lee 2006), and dextropropoxyphene plus aspirin (Bedi 1969). The active intervention in Brunnmuller 2004 was tilidine in combination with naloxone, and in the remaining two studies the active intervention was an opioid alone: pentazocine in Nuki 1973 and morphine sulphate in Moran Efficacy outcomes Five of the six studies reported superiority of the study drug compared with placebo on at least one efficacy measure (Bedi 1969; Boureau 1991; Brunnmuller 2004; Lee 2006; Moran 1991); no difference between pentazocine and placebo was found for any efficacy measure in Nuki None of the six studies reported the primary outcome measure of patient reported pain relief of 30% or greater. Boureau 1991 reported the proportion of participants achieving pain relief of 50% or greater in a seven-day trial of codeine and paracetamol versus placebo in patients with at least moderate pain due to RA: 12 of 20 (60%) participants who received the active drug achieved this outcome compared with 5 of 19 (26%) who received placebo (RR 2.28, 95% CI 0.99 to 5.25). There were insufficient continuous pain data in these six studies for a pooled analysis. Comparable measures of patient reported global impression of clinical change could be extracted from three studies (324 participants). Boureau 1991 and Lee 2006 reported the proportion of participants with self-reported global efficacy of good or very 14

17 good and Brunnmuller 2004 reported the proportion of participants who were improved versus not improved or worse. Although three different weak oral opioids were used in these studies (codeine with paracetamol, tramadol with paracetamol, tilidine with naloxone, respectively), the age and gender of participants, disease duration and background DMARD and NSAID use in each of the studies were sufficiently homogeneous to allow a meaningful pooling of results. The pooled analysis demonstrated opioids to be superior to placebo (RR 1.44, 95% CI 1.03 to 2.03). The estimated NNT for one extra patient to achieve a benefit rated as good or very good within the first six weeks was 6 (95% CI 3 to 84). An I 2 of 16% indicated a low degree of heterogeneity between trials. Four studies with 345 participants (Boureau 1991; Brunnmuller 2004; Lee 2006; Moran 1991) assessed the number of withdrawals due to inadequate analgesia. There was no significant difference between the opioid and placebo groups (RR 0.82, 95% CI 0.34 to 2.01). There was no statistical evidence for heterogeneity in the included trials (I 2 = 0%). Three studies assessed function. Lee 2006 and Moran 1991 measured the HAQ score at one week and two weeks, respectively. Neither study found a significant difference between groups, nor was there a difference in the pooled analysis (weighted MD -0.10, 95% CI to 0.13). Boureau 1991 used a self-reported disability scale (none, mild, moderate, severe) measured daily for seven days and found codeine plus paracetamol to be superior to placebo (repeated measures ANOVA, P = 0.04). Safety outcomes Data on the proportion of participants who experienced at least one adverse event could be pooled for four studies (three parallelgroup design, one cross-over study) with a total of 371 participants (Boureau 1991; Brunnmuller 2004; Lee 2006; Nuki 1973). Almost half (47.7%) of participants reported at least one adverse event, and the risk was significantly higher for those who received an opioid versus placebo (OR 3.90, 95% CI 2.31 to 6.56; I 2 = 0%). This equates to an estimated NNT with opioids to result in one additional adverse event within the first six weeks of opioid therapy of 4 (95% CI 3 to 6). A sensitivity analysis that excluded the cross-over study (Nuki 1973) was performed but the results were similar (OR 3.85, 95% CI 2.23 to 6.67). Adverse effects resulted in withdrawal from the trial for 14.8% of participants in the three parallel-group trials. The rate of withdrawal was higher for those who received an opioid, although the difference was not statistically significant (3 studies, 331 participants: RR 2.67, 95% CI 0.52 to 13.75). An I 2 of 57% indicated a moderate degree of heterogeneity between trials for this outcome. Data for withdrawals due to adverse events from the Nuki 1973 study could not be pooled, however adverse effects resulted in discontinuation of treatment in 3/35 participants during the pentazocine phase and 1/35 during the placebo phase. Another crossover study (Bedi 1969) reported no significant difference in adverse event frequency during treatment with aspirin alone or dextropropoxyphene plus aspirin, although 3/51 participants discontinued treatment due to adverse effects all during the combination therapy phase. Moran 1991 was the only study to use a strong opioid (controlled-release morphine sulphate). Adverse event data were not presented in a format that was suitable for meta-analysis but only 4/20 participants completed the trial: 11 participants withdrew during the first 4-week phase, 7 of whom were in the morphine group, and a further 5 participants withdrew in the second phase, 3 of whom were receiving morphine. Six of the 10 withdrawals during treatment with morphine were due to adverse effects compared with three of the six withdrawals during placebo treatment. The most commonly-reported adverse events in the 11 included studies were nausea and vomiting, dizziness or lightheadedness, and constipation. These data were insufficiently reported for a pooled analysis. One of the single-dose studies (Huskisson 1974) did not report any data on adverse events. Two serious adverse events (SAEs) were reported, one from each of two studies, although Lee 2006 was the only study to specify SAEs as an outcome measure. One participant in Lee 2006, in the tramadol plus paracetamol group, experienced chest pain, nausea and vomiting that were sufficiently severe to be classified as an SAE; the symptoms resolved within a day of discontinuation of the study drug. One participant in Bedi 1969, who was receiving dextropropoxyphene plus aspirin, developed haematemesis and melaena and was described as having an existing duodenal ulcer. The patient was withdrawn from the trial and no further details regarding the adverse event were given. Risk versus benefit In order to combine an estimate of both the benefits and harms of opioid therapy in RA into a single measure, a Net Efficacy Adjusted for Risk (NEAR) analysis was performed (Boada 2008). This method estimates the RR of simultaneously achieving a benefit and avoiding harm in study participants treated with opioid versus placebo since this state, a clinical benefit without a significant adverse drug reaction, is the optimal outcome in patients who require analgesic medications. For this analysis, a patient global impression of change of good or very good was chosen as the efficacy outcome, and adverse effects sufficient to warrant discontinuation of the drug were chosen as the most relevant measure of harm. Data for this analysis could be extracted from three studies (Boureau 1991; Brunnmuller 2004; Lee 2006), which included 324 participants. The NEAR RR (that is the risk of achieving benefit without harm for those treated with opioids versus placebo) was 1.20 (95% CI 0.89 to 1.61); therefore, by this measure, opioids were not superior to placebo. See Figure 3. 15

18 Figure 3. Net efficacy adjusted for risk Opioid versus NSAID One study (Glowinski 1999) with 60 participants compared opioid therapy with an NSAID. This one-week trial randomized participants on stable background therapy for RA to treatment with diclofenac 50 mg twice daily or one tablet of codeine 30 mg with paracetamol 500 mg three times daily plus diclofenac 50 mg at night. No difference was found between treatment groups in either efficacy outcomes (change in pain VAS, patient global impression, disability) or adverse effects. D I S C U S S I O N Summary of main results None of the 11 included studies reported the primary efficacy outcome of this review, patient reported pain relief of 30% or greater, and the secondary efficacy outcomes were variably reported. Four of the studies investigated the effectiveness of single doses of opioids; of the remaining seven studies the median duration was one week and the longest follow-up period was only six weeks. In general, the risk of bias in the included studies was high as the method by which treatment assignment was randomized and concealed was poorly reported, and participants with missing data were often excluded from analysis even in studies claiming to have performed an intention-to-treat analysis. The single-dose studies all reported superior pain relief with opioids compared with placebo but did not show superiority to non-opioid analgesics. Of the six studies that compared regular opioid therapy with placebo, five demonstrated superiority of opioids on at least one efficacy measure. No difference between treatments was found in the one study that compared an opioid with an NSAID. Dichotomous responder outcomes have been suggested to be the preferred outcome measures in pain trials due to the difficulty in interpreting the clinical relevance of group means for continuous outcomes when the underlying distribution is often skewed (Moore 2010a). Of the three dichotomous secondary outcome measures for efficacy planned for this review, none of the included studies reported the proportion of patients achieving a pain score below 30/100 mm on a visual analogue scale, and only one (Boureau 1991) reported patient reported pain relief of 50% or greater. The third, patient reported global impression of clinical change (PGIC) much or very much improved, or a comparable patient-reported global measure, was reported by three studies (Boureau 1991; Brunnmuller 2004; Lee 2006) and on pooled analysis opioids were found to be superior to placebo. Four studies reported the number of participant withdrawals due to inadequate analgesia; no difference between opioid and placebo was found for 16

19 this outcome in any of the studies or on pooled analysis. Only two studies reported functional outcomes and in neither case was there a difference between opioids and placebo. None of the included studies reported quality of life data. Despite the short duration of the included studies, approximately half of the participants who received an opioid reported at least one adverse event, and for one in six of these participants the adverse effects were sufficient to cause withdrawal from the study. Adverse events were more frequent among those treated with opioids in all of the placebo-controlled studies that were at least one week in duration, and pooled analysis found a four-fold increase in the odds of adverse events in opioid-treated patients. No difference in the incidence of adverse events was found in Bedi 1969, a crossover trial of dextropropoxyphene and aspirin versus aspirin alone, and Glowinski 1999, the only trial to directly compare an opioid with an NSAID. Only two serious adverse events were reported and there were no deaths. Where reported, the adverse events appeared to be relatively mild and were consistent with recognized opioid adverse effects, including nausea, vomiting, constipation and dizziness. A pooled analysis of withdrawals due to adverse effects in the placebo-controlled trials also suggested an increased risk in opioid-treated participants, although this difference was not statistically significant. After adjustment for adverse effects (the net efficacy adjusted for risk analysis), the superiority of opioids in achieving a PGIC of good or very good was no longer statistically significant suggesting that in patients with RA who are treated with opioids for up to six weeks, the analgesic benefits may be offset by adverse effects. Overall completeness and applicability of evidence The overall balance of benefits and harms associated with opioid use is a key issue for RA sufferers and the medical practitioners who care for them. RA is a chronic disease and analgesic medications may be required both in the short term (for example during flares of inflammatory activity or while waiting for DMARDs to take effect) and in the long term (for example due to irreversible joint damage or the development of secondary central pain syndromes such as fibromyalgia). It is not necessarily the case that the efficacy or safety of opioids, or the balance between the two, is similar for short-term versus long-term use or for pain of different mechanistic origin. The studies in this review were performed between 1969 and 2006, an era that has seen revolutionary changes in the management of the inflammatory component of RA. While none of the included studies attempted to describe the nature of the pain suffered by the participants, it is quite likely that pain of an inflammatory nature was more prominent in the earlier studies than those performed in the last decade. Although RA disease activity was not reported in the included studies, and DMARD use was incompletely reported, the use of DMARDs by participants in the included studies tended to be more prevalent in the latter studies. None of the participants were receiving biologic therapy. The significant changes in the management of RA that have occurred in the years covered by the studies in this review may limit the applicability of the results to RA patients in the current era, however pain remains a significant management challenge in RA today and the review highlights the need for further well-designed pain research in RA patients who are receiving modern diseasemodifying treatment strategies. For RA patients who experience persistent pain despite optimal control of synovial inflammation, there may be a long-term requirement for analgesic medications. There is a relative paucity of evidence regarding the risk-benefit profile of opioid analgesics in this population and in the longer term (Furlan 2006). Some unwanted effects, such as endocrinopathy and opioid-induced hyperalgesia, may increase in prevalence with chronic use, while tolerance to other adverse effects may develop (Chu 2008; Rhodin 2010). The risk of opioid addiction among long-term users is also a concern for many prescribers. A Cochrane review of long-term opioid use (at least six months) for chronic non-cancer pain (Noble 2010) found the rate of withdrawal due to adverse effects to be 11.4% (95% CI 7.0% to 8.1%) for weak oral opioids and 34.1% (95% CI 29.2% to 39.3%) for strong oral opioids. The rate of addiction or abuse was estimated at 0.27%. Clinically-relevant analgesia was achieved with oral opioids, however the included studies were predominantly uncontrolled and reliable effect sizes could not be estimated. Although none of the studies included in that review involved participants with RA, the conclusions are likely to be applicable to the RA population. High-quality trials of long-term opioid use in the setting of inflammatory arthritis and other non-malignant chronic pain states are clearly required. Quality of the evidence The 11 studies included in this review were heterogeneous with regard to the opioid studied, the concurrent use of non-opioid analgesics, the outcomes measured, the background disease-modifying anti-rheumatic therapy and the era in which the study was performed; and the risk of bias was generally high. In light of this, the quantitative findings of this review must be interpreted with great caution. At best, there is weak evidence in favour of the efficacy of opioids for the treatment of pain in patients with RA but, as no study was longer than six weeks in duration, no reliable conclusions can be drawn regarding the efficacy or safety of opioids in the longer term. Potential biases in the review process We believe that all relevant studies were identified. A thorough search strategy was devised and all major databases were searched for relevant studies with no language restrictions applied. Two 17

20 review authors assessed the trials for inclusion in the review and risk of bias, with a third review author adjudicating when there was any discrepancy. The biggest limitation of the review process was that many trials did not provide enough published data, or data in a form that could be extracted for meta-analysis. Agreements and disagreements with other studies or reviews No previous reviews of opioids in RA were identified. Findings from systematic reviews of opioid therapy in other chronic painful musculoskeletal syndromes may be informative, however. A Cochrane review of the treatment of chronic non-cancer pain with at least six months of opioid therapy (Noble 2010), discussed above, did not include any studies of RA but did include other painful musculoskeletal conditions, such as osteoarthritis and back pain. The conclusions were similar, that a proportion of patients with chronic non-malignant pain who are treated with opioids achieve clinically-significant analgesia, but adverse effects are common and limit use in many patients. Similarly, a Cochrane review of oral opioids (including codeine, oxycodone, oxymorphone and morphine) or transdermal fentanyl for osteoarthritis of the knee or hip (Nuesch 2009) found a mild improvement in pain scores with opioids versus control interventions but this was potentially outweighed by an increase in the risk of adverse events (RR 1.55, 95% CI 1.41 to 1.70) or withdrawal due to adverse events (RR 4.05, 95% CI 3.06 to 5.38). The authors concluded that, due to the poor risk-benefit profile, these opioids should not be routinely used for pain due to osteoarthritis of the knee or hip. A systematic review of the prevalence of adverse events in RCTs of opioids for chronic non-malignant pain (including, but not limited to, RA) found a rate of adverse events (approximately one in two participants) and withdrawal due to adverse events (approximately one in five) that were very similar to the findings of our review (Moore 2005). A U T H O R S C O N C L U S I O N S Implications for practice Based on the results of this review, treatment of RA patients with weak oral opioids for up to six weeks may offer clinically-relevant improvement in pain, but adverse effects are common and limit the utility of this class of analgesics. There is insufficient evidence to draw conclusions regarding the regular use of weak oral opioids for longer than six weeks, or the use of strong opioids for any duration. There is no existing evidence in this population that compares different opioids, different routes of delivery, or different dosing regimens. Implications for research More evidence regarding the effectiveness and safety of chronic opioid use in patients with RA and other inflammatory rheumatic diseases is required. Future RCTs should be adequately powered and should have follow-up periods of at least six weeks (and preferably 12 weeks or longer), clinically-relevant pain outcomes, rigorous collection of adverse event data, and blinded participants and outcome assessment. Further trials should seek to compare the risk-benefit profiles of different opioids, different routes of administration and combination analgesic strategies. Clinical effectiveness studies (described in Moore 2010b) that are structured to more accurately reflect decision-making in real-world management of chronic painful conditions would be of great value. The optimal management of pain in patients with RA remains a critical field of inquiry. A C K N O W L E D G E M E N T S The authors thank Louise Falzon, Trials Search Coordinator of the Cochrane Musculoskeletal Group, for assisting with the design of the search strategy. R E F E R E N C E S References to studies included in this review Bedi 1969 {published data only} Bedi SS. Comparison of aspirin and dextropropoxyphenewith-aspirin as analgesics in rheumatoid arthritis. The British Journal of Clinical Practice 1969;23(10): Boureau 1991 {published data only} Boureau F, Boccard E. Placebo-controlled study of the analgesic efficacy of a combination of paracetamol and codeine in rheumatoid arthritis. Acta Therapeutica 1991;17 (2): Brunnmuller 2004 {published data only} Brunnmuller U, Zeidler H, Alten R, Gromnica-Ihle E. Effective analgesic therapy with tilidine/naloxone for patients with rheumatoid arthritis. [German]. Aktuelle Rheumatologie 2004;29(1):35 9. Glowinski 1999 {published data only} Glowinski J, Boccard E. Placebo-controlled study of the analgesic efficacy of a paracetamol 500 mg/codeine 30 mg combination together with low-dose vs high-dose diclofenac in rheumatoid arthritis. Clinical Drug Investigation 1999;18 (3):