Advertisement

Long-Term Opioid Therapy for Chronic Noncancer Pain: A Systematic Review and Meta-Analysis of Efficacy and Safety

Open ArchivePublished:January 07, 2008DOI:https://doi.org/10.1016/j.jpainsymman.2007.03.015

      Abstract

      Opioid therapy for chronic noncancer pain (CNCP) is controversial due to concerns regarding long-term efficacy and adverse events (including addiction). We systematically reviewed the clinical evidence on patients treated with opioids for CNCP for at least six months. Of 115 studies identified by our search of eleven databases (through April 7, 2007), 17 studies (patients [n]=3,079) met inclusion criteria. Studies evaluated oral (studies [k]=7; n=1,504), transdermal (k=3; n=1, 993), and/or intrathecal (k=8; n=177) opioids. Many patients withdrew from the clinical trials due to adverse effects (oral: 32.5% [95% confidence interval (CI), 26.1%–39.6%]; intrathecal: 6.3% [95% CI, 2.9%–13.1%]; transdermal: 17.5% [95% CI, 6.5%–39.0%]), or due to insufficient pain relief (oral: 11.9% [95% CI, 7.8%–17.7%]; intrathecal: 10.5% [95% CI, 3.5%–27.4%]; transdermal: 5.8% [95% CI, 4.2%–7.3%]). Signs of opioid addiction were reported in only 0.05% (1/2,042) of patients and abuse in only 0.43% (3/685). There was an insufficient amount of data on transdermal opioids to quantify pain relief. For patients able to remain on oral or intrathecal opioids for at least six months, pain scores were reduced long-term (oral: standardized mean difference [SMD] 1.99, 95% CI, 1.17–2.80; intrathecal: SMD 1.33, 95% CI, 0.97–1.69). We conclude that many patients discontinue long-term opioid therapy due to adverse events or insufficient pain relief; however, weak evidence suggests that oral and intrathecal opioids reduce pain long-term in the relatively small proportion of individuals with CNCP who continue treatment.

      Key Words

      Background

      Chronic pain is an important and common medical concern worldwide. In the United States, pain is the most common complaint that leads patients to seek medical care. A systematic review of four international studies conducted in developed countries found prevalence rates of any type and severity level of chronic pain ranging from 10.5% to 55.2% of the population.
      • Harstall C.
      • Ospina M.
      How prevalent is chronic pain?.
      The Pain in Europe survey of 46,000 individuals showed that one in five people reported suffering from chronic pain. In this survey, chronic pain sufferers reported seven years of chronic pain on average, with some reporting pain lasting more than 20 years.
      • Breivik H.
      • Collett B.
      • Ventafridda V.
      • Cohen R.
      • Gallacher D.
      Survey of chronic pain in Europe: prevalence, impact on daily life, and treatment.
      A World Health Organization survey of primary care patients seeking care at 15 centers in 14 countries across Asia, Africa, Europe, South America, and North America found that 22% of primary care patients reported pain lasting more than six months.
      • Gureje O.
      • Korff Von
      • Simon G.E.
      • Gater R.
      Persistent pain and well-being: a World Health Organization Study in Primary Care.
      An estimated 9% of Americans
      • Clark J.D.
      Chronic pain prevalence and analgesic prescribing in a general medical population.
      and 19% of Europeans
      • Breivik H.
      • Collett B.
      • Ventafridda V.
      • Cohen R.
      • Gallacher D.
      Survey of chronic pain in Europe: prevalence, impact on daily life, and treatment.
      have moderate-to-severe chronic noncancer pain (CNCP). Older people and women are more likely to experience chronic pain.
      • Breivik H.
      • Collett B.
      • Ventafridda V.
      • Cohen R.
      • Gallacher D.
      Survey of chronic pain in Europe: prevalence, impact on daily life, and treatment.
      • Blyth F.M.
      • March L.M.
      • Brnabic A.J.
      • et al.
      Chronic pain in Australia: a prevalence study.
      • Elliott A.M.
      • Smith B.H.
      • Penny K.I.
      • Smith W.C.
      • Chambers W.A.
      The epidemiology of chronic pain in the community.
      • Bowsher D.
      • Rigge M.
      • Sopp L.
      Prevalence of chronic pain in the British population: a telephone survey of 1037 households.
      Although opioid use for acute/postsurgical pain and for palliative care is accepted in the United States, there is debate about whether opioids are appropriate for the treatment of CNCP,
      • American Academy of Pain Medicine, American Pain Society
      The use of opioids for the treatment of chronic pain. A consensus statement from the American Academy of Pain Medicine and the American Pain Society.
      • Jovey R.D.
      • Ennis J.
      • Gardner-Nix J.
      • et al.
      Use of opioid analgesics for the treatment of chronic noncancer pain—a consensus statement and guidelines from the Canadian Pain Society, 2002.
      although acceptance may be growing.
      • Kalso E.
      • Allan L.
      • Dellemijn P.L.
      • et al.
      Recommendations for using opioids in chronic non-cancer pain.
      The most commonly prescribed opioids in the United States are codeine, fentanyl, hydrocodone, hydromorphone, levorphanol, meperidine, methadone, morphine, and oxycodone.
      • Ballantyne J.C.
      • Mao J.
      Opioid therapy for chronic pain.
      The efficacy of opioids for CNCP has been demonstrated in short-term trials,
      • Furlan A.D.
      • Sandoval J.A.
      • Mailis-Gagnon A.
      • Tunks E.
      Opioids for chronic noncancer pain: a meta-analysis of effectiveness and side effects.
      including for neuropathic pain,
      • Eisenberg E.
      • McNicol E.
      • Carr D.B.
      Opioids for neuropathic pain.
      but little is known about whether these agents continue to be effective over the many months or years typical of CNCP. Concerns have also been raised about adverse effects that may arise with long-term use, including the development of psychological addiction and/or abuse. The purpose of this systematic review is to summarize the evidence pertaining to the efficacy and safety of long-term opioid therapy for CNCP.

      Methods

      Search Strategy

      Eleven databases were searched, including EMBASE (1980 through April 7, 2007) and PubMed (1966 through April 7, 2007) and all Cochrane databases and registries (inception through Issue 2, 2007). We used search terms including the following: chronic pain (pain, intractable [major heading(mh)]) OR (pain AND (chronic OR intractable OR refractory OR persistent OR chronic disease[mh]), a list of painful conditions, and all opioids in current use by generic and proprietary names. Full search strategies are available upon request from the contact author. We also examined reference lists from identified studies and reviewed gray literature for additional studies not identified by other means.

      Selection of Studies

      Systematic reviews aim to identify the best available evidence, which is often in the form of randomized controlled trials (RCTs). However, no placebo-controlled, long-term RCTs on the efficacy and safety of opioids for CNCP exist, and only one long-term controlled trial (which compares two opioids)
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      exists. Currently, the best available long-term evidence is from open-label, uncontrolled, time-series studies.
      Open-label, uncontrolled, time-series studies are susceptible to bias. Nonetheless, we chose to analyze data from these studies because (1) the patients in these studies had pain refractory to treatment for at least three months before enrollment (i.e., International Association for the Study of Pain [IASP] definition); and (2) all patients had failed previous pharmacotherapy before beginning opioids (in accordance with the World Health Organization [WHO] analgesic ladder). These factors suggest that patients' pain levels probably would not have lessened without active treatment.
      We searched for studies that (1) collected data on patients after at least six months of opioid therapy; (2) were published in English; (3) were reported as full-text articles; (4) did not include patients also reported on in other included studies; (5) were prospective; (6) enrolled at least 10 patients; (7) enrolled only patients who had CNCP, defined as pain lasting at least three months as defined by IASP. We used two additional criteria for pain outcomes: (1) pain outcomes must have been patient-reported; and (2) outcome data must not have been collected retrospectively (e.g., post-treatment surveys/questionnaires), because reports based on memory of pain may differ from reports given at the time that pain is experienced.
      • Eich E.
      • Reeves J.L.
      • Jaeger B.
      • Graff-Radford S.B.
      Memory for pain: relation between past and present pain intensity.
      • Linton S.J.
      • Gotestam K.G.
      A clinical comparison of two pain scales: correlation, remembering chronic pain, and a measure of compliance.

      Data Collection

      Data on patient withdrawal from the study due to adverse events/effects, the incidence and type of adverse drug effects (including addiction and abuse rates), withdrawal due to insufficient pain relief and pain relief were independently extracted by two researchers. Discrepancies were settled by consensus.

      Analysis

      Two methodologists independently assessed the quality of all identified studies using a 14-item quality assessment instrument (Table 1). This quality assessment scale was developed by ECRI Institute to assess the internal validity of studies using domains identified as important factors by experts in the field.
      • Agency for Healthcare Research and Quality (AHRQ)
      Systems to rate the strength of scientific evidence: Summary.
      • Egger M.
      • Juni P.
      • Bartlett C.
      • Holenstein F.
      • Sterne J.
      How important are comprehensive literature searches and the assessment of trial quality in systematic reviews? Empirical study.
      • Deeks J.J.
      • Dinnes J.
      • D'Amico R.
      • et al.
      Evaluating non-randomised intervention studies.
      We assessed the quality of each included study for each outcome of interest and used the median quality of the studies for each outcome to describe the overall evidence-base quality (refer to Table 1).
      Table 1ECRI Institute Before–After Study Quality Scale
      Quality Assessment Item
      1. Does the study specifically state that previous treatments for this disease, disorder, or condition had been attempted and were unsuccessful?
      2. Was the study prospective?
      3. Did the study enroll ALL patients or CONSECUTIVE patients?
      4. Were the criteria for including and excluding patients based on objective laboratory and/or clinical findings?
      5. Were the patient inclusion/exclusion criteria established a priori?
      6. Was the same treatment given to all patients enrolled? (If treatment was specifically tailored to individual patients, and was reported by authors to be in the same category of treatment, answer “Yes.”)
      7. Was the study free of confounding interventions?
      8. Were the outcome measures objective?
      9. Did ≥85% of patients complete the study?
      10. Were the characteristics of those who did and did not complete the study compared, and were these characteristics similar?
      11. Was the analysis conducted on an intent-to-treat basis, or was data presented in such a way that one could be performed?
      12. Did the study have at least 80% statistical power to detect a clinically important change in the outcome variable after the intervention?
      13. Was the funding for this study derived from a source that does not have a financial interest in its results?
      14. Were the author's conclusions, as stated in the abstract or the article's discussion section, supported by the data presented in the article's results section?
      We analyzed the available evidence using systematic a priori protocols.
      • Treadwell J.T.
      • Tregear S.J.
      • Reston J.T.
      • Turkelson C.M.
      A system for rating the stability and strength of medical evidence.
      We pooled data using meta-analysis only when at least three studies per mode of opioid administration addressed a particular outcome of interest. These protocols were used to formulate both qualitative (i.e., the direction of a treatment effect) and quantitative (i.e., the size of a treatment effect) conclusions.
      Follow-up times in included studies ranged from 6 to 48 months. We did not pool data for specific time points, but chose instead to use the last reported time point of each study and then performed meta-regression to assess whether the treatment duration affected the outcomes. We took this approach because too few studies reported data at any given time point to enable multiple time point analyses. Reporting of data in intrathecal studies differed from oral and transdermal opioid studies in that means of results from patients' last reported follow-up visits were presented as results for the mean length of follow-up.
      Fixed-effects and random-effects meta-analyses were used to pool data from different studies.
      • Hedges L.V.
      Fixed effects models.
      • Raudenbush S.W.
      Random effects models.
      • Shadish W.R.
      • Haddock C.K.
      Combining estimates of effect size.
      • Hedges L.V.
      • Vevea J.L.
      Fixed- and random-effects models in meta-analysis.
      We used the I2-statistic to quantify the magnitude of important differences in the findings of different studies (i.e., heterogeneity).
      • Greenhouse J.B.
      • Iyengar S.
      Sensitivity analysis and diagnostics.
      • Higgins J.P.
      • Thompson S.G.
      • Deeks J.J.
      • Altman D.G.
      Measuring inconsistency in meta-analyses.
      When no substantial heterogeneity (I250%) was detected, we used the fixed-effects model. Whenever appropriate, heterogeneity was explored with meta-regression techniques using a permutation test.
      • Thompson S.G.
      • Higgins J.P.
      How should meta-regression analyses be undertaken and interpreted?.
      • Higgins J.P.
      • Thompson S.G.
      Controlling the risk of spurious findings from meta-regression.
      We decided a priori to investigate duration of treatment as a covariate whenever we detected substantial heterogeneity, given our assumption that length of treatment would be of particular importance. We acknowledge, however, that the small number of studies limits the power to detect an effect. For a meta-regression using patient characteristics such as age, diagnosis, or duration of pain, patient-level data would be preferable, because it would increase statistical power and would be less susceptible to the ecological fallacy. However, such data were not available to this project. Therefore, we decided against exploring additional possible sources of heterogeneity (e.g., type of pain, patient characteristics, type and dosing of opioid, degree of flexibility in dosing, concomitant therapies, and use of preventive treatment for common adverse events). When heterogeneity was present that could not be explained by duration of follow-up, we used the random effects model for meta-analysis.
      We calculated the standardized mean difference (SMD) for continuous and categorical data (e.g., visual analog scale [VAS] pain scores and Likert scales). Because time series data are not independent, SMDs were adjusted by assuming a correlation of 0.50. If means and standard deviations were not available for continuous data, every effort was made to determine an estimate of treatment effect from reported statistics (e.g., t-values, f-values) or from P-values using methods described in detail elsewhere.
      Event rates (odds) were calculated for dichotomous data. Results are presented as percentages or numeric pain scores and 95% confidence intervals (CIs).
      Cumulative meta-analysis
      • Lau J.
      • Schmid C.H.
      • Chalmers T.C.
      Cumulative meta-analysis of clinical trials builds evidence for exemplary medical care.
      • Ioannidis J.P.
      • Contopoulos-Ioannidis D.G.
      • Lau J.
      Recursive cumulative meta-analysis: a diagnostic for the evolution of total randomized evidence from group and individual patient data.
      • Conti C.R.
      Clinical decision making using cumulative meta-analysis.
      and impact meta-analysis (in which we removed and replaced one study at a time) were performed for all outcomes. For continuous outcomes, assumptions about correlation coefficients were tested by sensitivity analyses in which we varied the correlation coefficient from 0.00 to 1.00 to determine whether the summary effect size was stable, in the method of Thiessen Philbrook and colleagues.
      • Thiessen Philbrook H.
      • Barrowman N.
      • Garg A.X.
      Imputing variance estimates do not alter the conclusions of a meta-analysis with continuous outcomes: a case study of changes in renal function after living kidney donation.
      Because patients who withdraw may differ in important ways from those who remain in a study, we also performed sensitivity analyses in which we inflated baseline standard deviations for pain scores by 100% and 200%. We considered more than a 25% change in summary effect size upon manipulation of the correlation coefficient and/or the baseline standard deviations to indicate that the quantitative estimate of the meta-analysis was not stable. For all fixed-effects meta-analyses, we also computed a random effect meta-analysis in addition to a fixed-effect meta-analysis, to determine whether the type of meta-analysis model used substantially affected the summary effect size and/or CIs. In evidence bases without substantial heterogeneity (I250%) we used the “trim-and-fill” method to test for funnel plot asymmetry, which suggests missing studies, possibly due to publication bias.
      • Duval S.
      • Tweedie R.
      Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis.
      • Sutton A.J.
      • Duval S.J.
      • Tweedie R.L.
      • Abrams K.R.
      • Jones D.R.
      Empirical assessment of effect of publication bias on meta-analyses.
      • Duval S.J.
      • Tweedie R.L.
      A non-parametric ‘trim and fill’ method of assessing publication bias in meta-analysis.
      Comprehensive Meta-Analysis (Biostat, Englewood, NJ) software was used for most statistical analyses. Meta-regression with permutation tests was performed using STATA (StataCorp LP, Bryan/College Station, TX) software.

      Findings

      Study Selection

      Our searches identified 3,672 abstracts from which we retrieved 115 studies; 17 of those studies met our inclusion criteria (Fig. 1). All but one of the 17 studies were open-label, single-arm time series. The other study
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      included one group in which patients were treated with transdermal fentanyl and another in which patients were treated with oral extended-release morphine.

      Study Description

      All of the evidence bases considered in this systematic review were of low quality. Their low quality was primarily due to failure to describe patient recruitment methods, failure to compare the characteristics of patients who completed the study to patients who did not, high patient withdrawal rates, and failure to conduct intent-to-treat (ITT) analyses in addition to completer analyses. For pain outcomes, studies met 8.2 quality criteria on average (range 7–10). For withdrawal outcomes, studies met an average of 9.5 quality criteria (range 9–11) (see Table 1).
      Time-series data for 18 separate treatment groups from 17 separate studies that enrolled a total of 3,808 patients with CNCP formed the complete evidence base for this review (Table 2). This evidence base was stratified by mode of opioid administration: oral (7 treatment groups, n=1,504), transdermal (3 treatment groups, n=1,399), and intrathecal pump (8 treatment groups, n=177). Patients in all treatment groups reported moderate-to-severe pain at baseline due to nociceptive or neuropathic pain, or both. Types of pain included general or mixed CNCP (studies [k]=2),
      • Milligan K.
      • Lanteri-Minet M.
      • Borchert K.
      • et al.
      Evaluation of long-term efficacy and safety of transdermal fentanyl in the treatment of chronic noncancer pain.
      • Mystakidou K.
      • Parpa E.
      • Tsilika E.
      • et al.
      Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
      neuropathic pain (k=2),
      • Hassenbusch S.J.
      • Stanton-Hicks M.
      • Covington E.C.
      • Walsh J.G.
      • Guthrey D.S.
      Long-term intraspinal infusions of opioids in the treatment of neuropathic pain.
      • Harati Y.
      • Gooch C.
      • Swenson M.
      • et al.
      Maintenance of the long-term effectiveness of tramadol in treatment of the pain of diabetic neuropathy.
      osteoarthritis (k=3),
      • Caldwell J.R.
      • Rapoport R.J.
      • Davis J.C.
      • et al.
      Efficacy and safety of a once-daily morphine formulation in chronic, moderate-to-severe osteoarthritis pain: results from a randomized, placebo-controlled, double-blind trial and an open-label extension trial.
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      • Peloso P.M.
      • Bellamy N.
      • Bensen W.
      • et al.
      Double blind randomized placebo control trial of controlled release codeine in the treatment of osteoarthritis of the hip or knee.
      back pain (k=8),
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      • Anderson V.C.
      • Burchiel K.J.
      A prospective study of long-term intrathecal morphine in the management of chronic nonmalignant pain.
      • Anderson V.C.
      • Burchiel K.J.
      • Cooke B.
      A prospective, randomized trial of intrathecal injection vs. epidural infusion in the selection of patients for continuous intrathecal opioid therapy.
      • Angel I.F.
      • Gould Jr., H.J.
      • Carey M.E.
      Intrathecal morphine pump as a treatment option in chronic pain of nonmalignant origin.
      • Fredheim O.M.
      • Kaasa S.
      • Dale O.
      • et al.
      Opioid switching from oral slow release morphine to oral methadone may improve pain control in chronic non-malignant pain: a nine-month follow-up study.
      • Kanoff R.B.
      Intraspinal delivery of opiates by an implantable, programmable pump in patients with chronic, intractable pain of nonmalignant origin.
      • Mironer Y.E.
      • Tollison C.D.
      Methadone in the intrathecal treatment of chronic nonmalignant pain resistant to other neuroaxial agents: the first experience.
      • Rainov N.G.
      • Heidecke V.
      • Burkert W.
      Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
      low back and leg pain (k=1),
      • Tutak U.
      • Doleys D.M.
      Intrathecal infusion systems for treatment of chronic low back and leg pain of noncancer origin.
      and neuropathic and back pain (k=1).
      • Zenz M.
      • Strumpf M.
      • Tryba M.
      Long-term oral opioid therapy in patients with chronic nonmalignant pain.
      Across all included studies, the mean ages of these patients ranged from 44
      • Kanoff R.B.
      Intraspinal delivery of opiates by an implantable, programmable pump in patients with chronic, intractable pain of nonmalignant origin.
      to 60.2
      • Fredheim O.M.
      • Kaasa S.
      • Dale O.
      • et al.
      Opioid switching from oral slow release morphine to oral methadone may improve pain control in chronic non-malignant pain: a nine-month follow-up study.
      years. The percentage of females enrolled ranged from 35%
      • Tutak U.
      • Doleys D.M.
      Intrathecal infusion systems for treatment of chronic low back and leg pain of noncancer origin.
      to 62.5%.
      • Mironer Y.E.
      • Tollison C.D.
      Methadone in the intrathecal treatment of chronic nonmalignant pain resistant to other neuroaxial agents: the first experience.
      The duration of pain prior to study enrollment was reported by only five studies, but where reported, ranged from 19 months
      • Rainov N.G.
      • Heidecke V.
      • Burkert W.
      Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
      to 10.3 years.
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      Table 2Characteristics of Included Studies
      Mode of AdministrationReferenceOpioidPredominant CNCP Type(s)n EnrolledOutcomes Used in Evidence Synthesis
      Withdrawal Due to Adverse EventsWithdrawal Due to Insufficient Pain ReliefPain
      Continuous/Categorical≥30% Relief≥50% Relief
      OralAllan et al.
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      Morphine
      Sustained release.
      Low back pain342
      Not analyzed because number of patients at follow-up times not reported.
      Caldwell et al.
      • Caldwell J.R.
      • Rapoport R.J.
      • Davis J.C.
      • et al.
      Efficacy and safety of a once-daily morphine formulation in chronic, moderate-to-severe osteoarthritis pain: results from a randomized, placebo-controlled, double-blind trial and an open-label extension trial.
      Morphine
      Sustained release.
      Osteoarthritis(295)
      N in parentheses denotes number of patients randomized in original RCT; second number is that enrolled in open-label extension.
      181
      Not meta-analyzed because reported units are statistically incompatible with the three other studies meeting inclusion criteria.
      Harati et al.
      • Harati Y.
      • Gooch C.
      • Swenson M.
      • et al.
      Maintenance of the long-term effectiveness of tramadol in treatment of the pain of diabetic neuropathy.
      TramadolDiabetic neuropathy(131) 117
      Fredheim et al.
      • Fredheim O.M.
      • Kaasa S.
      • Dale O.
      • et al.
      Opioid switching from oral slow release morphine to oral methadone may improve pain control in chronic non-malignant pain: a nine-month follow-up study.
      MethadoneLow back pain12
      Not analyzed because data were reported for fewer than ten patients at follow-up times.
      McIlwain and Ahdieh
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      Extended-release oxymorphoneOsteoarthritis(491) 153
      Roth et al.
      • Roth S.H.
      • Fleischmann R.M.
      • Burch F.X.
      • et al.
      Around-the-clock, controlled-release oxycodone therapy for osteoarthritis-related pain: placebo-controlled trial and long-term evaluation.
      Controlled-release oxycodoneOsteoarthritis(133) 106
      Zenz et al.
      • Zenz M.
      • Strumpf M.
      • Tryba M.
      Long-term oral opioid therapy in patients with chronic nonmalignant pain.
      Dihydrocodeine,
      Sustained release.
      buprenorphine, or morphine
      Sustained release.
      Neuropathic or back pain100
      TransdermalAllan et al.
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      FentanylLow back pain338
      Not analyzed because number of patients at follow-up times not reported.
      Milligan et al.
      • Milligan K.
      • Lanteri-Minet M.
      • Borchert K.
      • et al.
      Evaluation of long-term efficacy and safety of transdermal fentanyl in the treatment of chronic noncancer pain.
      FentanylUnspecified532
      Not analyzed because instrument used not validated.
      Mystakidou et al.
      • Mystakidou K.
      • Parpa E.
      • Tsilika E.
      • et al.
      Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
      FentanylUnspecified529
      IntrathecalAnderson and Burchiel
      • Anderson V.C.
      • Burchiel K.J.
      A prospective study of long-term intrathecal morphine in the management of chronic nonmalignant pain.
      MorphineNociceptive and neuropathic30
      Anderson et al.
      • Anderson V.C.
      • Burchiel K.J.
      • Cooke B.
      A prospective, randomized trial of intrathecal injection vs. epidural infusion in the selection of patients for continuous intrathecal opioid therapy.
      MorphineFailed back surgery syndrome (FBSS)27
      Angel et al.
      • Angel I.F.
      • Gould Jr., H.J.
      • Carey M.E.
      Intrathecal morphine pump as a treatment option in chronic pain of nonmalignant origin.
      MorphineFailed back syndrome11
      Hassenbusch et al.
      • Hassenbusch S.J.
      • Stanton-Hicks M.
      • Covington E.C.
      • Walsh J.G.
      • Guthrey D.S.
      Long-term intraspinal infusions of opioids in the treatment of neuropathic pain.
      Morphine or sufentanil citrateNeuropathic18
      Kanoff
      • Kanoff R.B.
      Intraspinal delivery of opiates by an implantable, programmable pump in patients with chronic, intractable pain of nonmalignant origin.
      MorphineMixed15
      Mironer and Tollison
      • Mironer Y.E.
      • Tollison C.D.
      Methadone in the intrathecal treatment of chronic nonmalignant pain resistant to other neuroaxial agents: the first experience.
      MethadoneFBSS24
      Rainov et al.
      • Rainov N.G.
      • Heidecke V.
      • Burkert W.
      Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
      Morphine (with bupivacaine, clonidine, or midazolam)FBSS26
      Tutak and Doleys
      • Tutak U.
      • Doleys D.M.
      Intrathecal infusion systems for treatment of chronic low back and leg pain of noncancer origin.
      Morphine or fentanylUnspecified26
      a Sustained release.
      b Not analyzed because number of patients at follow-up times not reported.
      c N in parentheses denotes number of patients randomized in original RCT; second number is that enrolled in open-label extension.
      d Not meta-analyzed because reported units are statistically incompatible with the three other studies meeting inclusion criteria.
      e Not analyzed because data were reported for fewer than ten patients at follow-up times.
      f Not analyzed because instrument used not validated.
      Patient withdrawal rates were generally high in the included studies. For example, although the RCTs of oral opioids were short-term (six weeks or less),
      • Harati Y.
      • Gooch C.
      • Swenson M.
      • et al.
      Maintenance of the long-term effectiveness of tramadol in treatment of the pain of diabetic neuropathy.
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      • Roth S.H.
      • Fleischmann R.M.
      • Burch F.X.
      • et al.
      Around-the-clock, controlled-release oxycodone therapy for osteoarthritis-related pain: placebo-controlled trial and long-term evaluation.
      drop-out rates were high and reduced the number of patients eligible to continue in the open-label phase.
      • Caldwell J.R.
      • Rapoport R.J.
      • Davis J.C.
      • et al.
      Efficacy and safety of a once-daily morphine formulation in chronic, moderate-to-severe osteoarthritis pain: results from a randomized, placebo-controlled, double-blind trial and an open-label extension trial.
      In the four studies of this type, 918 patients enrolled in the RCT portions of these studies, 450 (49.0%) participated in open-label extension studies, and 162 (17.6%) reported pain scores usable in the analysis (at follow-up times ranging from 6 to 18 months). At six months, withdrawal rates were lowest in intrathecal studies (range 0%
      • Hassenbusch S.J.
      • Stanton-Hicks M.
      • Covington E.C.
      • Walsh J.G.
      • Guthrey D.S.
      Long-term intraspinal infusions of opioids in the treatment of neuropathic pain.
      • Angel I.F.
      • Gould Jr., H.J.
      • Carey M.E.
      Intrathecal morphine pump as a treatment option in chronic pain of nonmalignant origin.
      • Mironer Y.E.
      • Tollison C.D.
      Methadone in the intrathecal treatment of chronic nonmalignant pain resistant to other neuroaxial agents: the first experience.
      • Rainov N.G.
      • Heidecke V.
      • Burkert W.
      Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
      • Tutak U.
      • Doleys D.M.
      Intrathecal infusion systems for treatment of chronic low back and leg pain of noncancer origin.
      to 17%
      • Anderson V.C.
      • Burchiel K.J.
      A prospective study of long-term intrathecal morphine in the management of chronic nonmalignant pain.
      ). Reported six-month withdrawal rates were higher in oral opioid studies (range 35%
      • Harati Y.
      • Gooch C.
      • Swenson M.
      • et al.
      Maintenance of the long-term effectiveness of tramadol in treatment of the pain of diabetic neuropathy.
      to 86%
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      ) and transdermal opioid studies (33%
      • Mystakidou K.
      • Parpa E.
      • Tsilika E.
      • et al.
      Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
      and 44%
      • Milligan K.
      • Lanteri-Minet M.
      • Borchert K.
      • et al.
      Evaluation of long-term efficacy and safety of transdermal fentanyl in the treatment of chronic noncancer pain.
      ). At one year, a similar trend was observed (oral: 52%
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      to 88%
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      ; transdermal: 43%
      • Milligan K.
      • Lanteri-Minet M.
      • Borchert K.
      • et al.
      Evaluation of long-term efficacy and safety of transdermal fentanyl in the treatment of chronic noncancer pain.
      to 67%
      • Mystakidou K.
      • Parpa E.
      • Tsilika E.
      • et al.
      Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
      ; intrathecal: 0%
      • Rainov N.G.
      • Heidecke V.
      • Burkert W.
      Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
      to 33%
      • Tutak U.
      • Doleys D.M.
      Intrathecal infusion systems for treatment of chronic low back and leg pain of noncancer origin.
      ). In intrathecal studies, withdrawal rates ranged from 0%
      • Rainov N.G.
      • Heidecke V.
      • Burkert W.
      Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
      to 33%
      • Anderson V.C.
      • Burchiel K.J.
      A prospective study of long-term intrathecal morphine in the management of chronic nonmalignant pain.
      at months 18 and 24, and from 61%
      • Hassenbusch S.J.
      • Stanton-Hicks M.
      • Covington E.C.
      • Walsh J.G.
      • Guthrey D.S.
      Long-term intraspinal infusions of opioids in the treatment of neuropathic pain.
      to 73%
      • Angel I.F.
      • Gould Jr., H.J.
      • Carey M.E.
      Intrathecal morphine pump as a treatment option in chronic pain of nonmalignant origin.
      at 36 months. The longest reported follow-up time of any of the intrathecal studies was 48 months, at which time 95% withdrawal was observed.
      • Hassenbusch S.J.
      • Stanton-Hicks M.
      • Covington E.C.
      • Walsh J.G.
      • Guthrey D.S.
      Long-term intraspinal infusions of opioids in the treatment of neuropathic pain.
      Only one study each of oral and transdermal administration presented data beyond one year, with both reporting 88% withdrawal at 18 months.
      • Mystakidou K.
      • Parpa E.
      • Tsilika E.
      • et al.
      Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
      • Roth S.H.
      • Fleischmann R.M.
      • Burch F.X.
      • et al.
      Around-the-clock, controlled-release oxycodone therapy for osteoarthritis-related pain: placebo-controlled trial and long-term evaluation.
      In the transdermal study, the withdrawal rate was 95% by 48 months (the longest reported follow-up time).
      • Mystakidou K.
      • Parpa E.
      • Tsilika E.
      • et al.
      Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
      Insufficient data were provided to statistically analyze prognostic factors for withdrawal for any reason.

      Evidence Synthesis

      The number of studies and enrolled patients in all analyses in this systematic review are presented in Table 2. Because patient withdrawal rates were so high, we first analyzed withdrawal rates for the two most common reasons cited for leaving a study: adverse events and insufficient pain relief. Therefore, the patients who reported long-term pain outcomes represent only a subset of the patients initially enrolled in the studies. For the four studies that were open-label extensions of RCTs, patients who withdrew from the active treatment group during the RCT portion were included in our analyses.
      • Harati Y.
      • Gooch C.
      • Swenson M.
      • et al.
      Maintenance of the long-term effectiveness of tramadol in treatment of the pain of diabetic neuropathy.
      • Caldwell J.R.
      • Rapoport R.J.
      • Davis J.C.
      • et al.
      Efficacy and safety of a once-daily morphine formulation in chronic, moderate-to-severe osteoarthritis pain: results from a randomized, placebo-controlled, double-blind trial and an open-label extension trial.
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      • Roth S.H.
      • Fleischmann R.M.
      • Burch F.X.
      • et al.
      Around-the-clock, controlled-release oxycodone therapy for osteoarthritis-related pain: placebo-controlled trial and long-term evaluation.

      Withdrawal from Clinical Study Due to Adverse Events

      Among individuals with CNCP taking oral opioids, approximately one-third did not continue long-term treatment (follow-up time range 6–18 months) because of intolerable adverse effects (32.5% [95% CI, 26.1%–39.6%]). Intrathecal studies reported the lowest drop out rates due to adverse events at longest follow up (6–mean of 29 months): 6.3% (95% CI, 2.9%–13.1%). A small amount of funnel plot asymmetry attributed to two missing studies was detected among the intrathecal studies. Adjusted by the trim-and-fill method, the event rate is slightly higher, 8.0% (95% CI, 4.1–15.2%). Although we did not consider this to be a substantial difference, we acknowledge the small number of studies analyzed (k=6) limits the power of the statistic to detect an effect. Transdermal trials reported intermediate adverse event-related drop-out rates at longest follow-up (range 12–48 months), at 17.5% (95% CI, 6.5%–39.0%). Forest plots of these findings are shown in Fig. 2. No substantial heterogeneity was detected among intrathecal studies (I2<0.001%); however, substantial heterogeneity was detected for both oral (I2=82.9%) and transdermal trials (I2=98.2%). In neither case could this heterogeneity be explained by differences across studies in the longest follow-up time, which ranged widely.
      Figure thumbnail gr2
      Fig. 2Withdrawal from clinical trial due to adverse events.

      Adverse Events

      Given the high rate of withdrawal from the clinical trials due to adverse events, we examined adverse events reported by all included studies. The variability in thresholds for reporting adverse events, failure to report the absence of unobserved but potential adverse events in some studies, inconsistent reporting or use of definitions of events/effects, and absence of control groups precluded our ability to pool data for meta-analysis.
      The most commonly reported adverse events included gastrointestinal effects (i.e., constipation, nausea, and dyspepsia), headache, fatigue/lethargy/somnolence, and urinary complications (i.e., retention, hesitancy, and “disturbance”). Adverse events unique to the intrathecal mode of administration included pump and catheter malfunctions and malpositioning, surgical complications, and postsurgical complications. The percentage of patients whose device complications required reoperation was as high as 35%.
      • Tutak U.
      • Doleys D.M.
      Intrathecal infusion systems for treatment of chronic low back and leg pain of noncancer origin.
      Of particular interest are addiction and abuse rates observed during long-term opioid treatment. Despite the importance of this issue, only seven of the 17 included studies specifically mentioned opioid addiction.
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      • Milligan K.
      • Lanteri-Minet M.
      • Borchert K.
      • et al.
      Evaluation of long-term efficacy and safety of transdermal fentanyl in the treatment of chronic noncancer pain.
      • Mystakidou K.
      • Parpa E.
      • Tsilika E.
      • et al.
      Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
      • Hassenbusch S.J.
      • Stanton-Hicks M.
      • Covington E.C.
      • Walsh J.G.
      • Guthrey D.S.
      Long-term intraspinal infusions of opioids in the treatment of neuropathic pain.
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      • Anderson V.C.
      • Burchiel K.J.
      A prospective study of long-term intrathecal morphine in the management of chronic nonmalignant pain.
      • Zenz M.
      • Strumpf M.
      • Tryba M.
      Long-term oral opioid therapy in patients with chronic nonmalignant pain.
      These studies included a total of 2,042 patients (not including an additional 259 patients in the original six-week RCT
      • Matsumoto A.K.
      • Babul N.
      • Ahdieh H.
      Oxymorphone extended-release tablets relieve moderate to severe pain and improve physical function in osteoarthritis: results of a randomized, double-blind, placebo- and active-controlled phase III trial.
      preceding the extension reported by McIlwain and Ahdieh,
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      for whom addiction rates were not reported). Reporting of addiction and abuse was inconsistent because not all studies reported this outcome. Among studies that did report addiction and abuse, which diagnostic criteria were used, and whether addiction/abuse monitoring was prospectively planned were generally unclear. Therefore, we included any data potentially suggesting addiction or abuse, although we recognize that this could lead to overestimation of actual addiction and/or abuse rates. Of the 2,042 patients in the 17 studies that reported signs of addiction and/or addictive behaviors, only one patient was reported as having possibly developed opioid addiction (in Anderson 1999,
      • Anderson V.C.
      • Burchiel K.J.
      A prospective study of long-term intrathecal morphine in the management of chronic nonmalignant pain.
      in the form of “drug-seeking behavior”). Therefore, in these seven studies the overall addiction rate was 0.042%. For the other 10 studies (n=680) that met general inclusion criteria, it is unclear whether addiction was not mentioned because it was not an outcome of interest, or because it was not observed. Milligan and colleagues
      • Milligan K.
      • Lanteri-Minet M.
      • Borchert K.
      • et al.
      Evaluation of long-term efficacy and safety of transdermal fentanyl in the treatment of chronic noncancer pain.
      reported three cases of opioid abuse (defined as taking opioids in inappropriate doses, or inappropriately with other drugs) among the 532 patients who enrolled in this 13-month study. Although drug abuse is not mentioned in the article describing their original six-week RCT (an additional 259 treated with opioids),
      • Matsumoto A.K.
      • Babul N.
      • Ahdieh H.
      Oxymorphone extended-release tablets relieve moderate to severe pain and improve physical function in osteoarthritis: results of a randomized, double-blind, placebo- and active-controlled phase III trial.
      McIlwain and Ahdieh reported that no opioid abuse was observed in any of the patients (n=153) enrolled in their 13-month, open-label extension study.
      • McIlwain H.
      • Ahdieh H.
      Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
      If the original RCT population from McIlwain and Ahdieh is not included, the total abuse rate from these two studies is 3/685 (0.43%). Given the widespread concern about abuse, it is unclear why it was not mentioned in the other 15 studies. If none occurred, the rate would be lower.

      Withdrawal from Clinical Study Due to Insufficient Pain Relief

      As with withdrawal from clinical studies due to adverse events, the highest summary event rate of discontinuation due to inadequate pain relief was observed in long-term (range 6–18 months) studies of oral opioids (11.9% [95% CI, 7.8%–17.7%]). The next highest summary discontinuation rate occurred among individuals enrolled in long-term (range 6 months to a mean of 29 months) intrathecal opioid studies (10.5% [95% CI, 3.5%–27.4%]). The lowest summary discontinuation rate due to insufficient pain control was observed in long-term (range 12–48 months) transdermal opioid studies; however, CIs overlap with those of oral opioids (5.8% [95% CI, 4.2%–7.3%]). Forest plots showing these effect sizes are shown in Fig. 3. As with withdrawal from clinical studies due to adverse events, substantial heterogeneity was not found for withdrawal due to insufficient pain relief among the studies of transdermal opioid use (I2=48.9%), but was detected in oral opioid (I2=85.5%) and intrathecal (I2=63.3%) studies. This heterogeneity could not be explained by differences across studies in the widely ranging longest follow-up times.
      Figure thumbnail gr3
      Fig. 3Withdrawal from clinical trial due to insufficient pain relief.

      Pain

      We could not formulate evidence-based conclusions on long-term efficacy of transdermal opioids for CNCP because only two studies contributed data to this analysis. For oral and intrathecal administration of opioids, data were sufficient for analyzing long-term pain score data. The included studies assessed pain severity with a visual analog scale (VAS) (0–10 or 0–100) or Likert scales (0–3 or 0–4). We combined the studies in meta-analysis and found that patients treated with long-term (range 6–mean of 29 months) intrathecal opioids had an approximately 38.0% mean reduction in pain scores (SMD=1.33 [95% CI, 0.97–1.69]), and patients treated with long-term (range 6–18 months) oral opioids had approximately a 63.4% mean reduction in pain scores (SMD=1.99 [95% CI, 1.17–2.80]). Substantial heterogeneity was observed in both oral studies (I2=86.6%) and intrathecal studies (I2=51.3%). The heterogeneity of oral and intrathecal opioid studies could not be resolved using meta-regression by follow-up time.
      The summary effect estimate of pain relief from oral and intrathecal opioids was not robust to sensitivity analyses manipulating baseline standard deviations, and the summary effect size of intrathecal opioids was also not robust to changes in the correlation coefficient. These findings suggest that the variability in pain scores of patients at baseline may affect the amount of pain relief attained. Due to this lack of robustness upon sensitivity analysis and unexplained heterogeneity, the quantitative estimates of the amount of pain relief associated with opioid therapy may be unstable. However, it was still evident that long-term opioids are associated with some degree of pain relief because even the smallest summary effect sizes yielded by the sensitivity analyses remained large (≥0.853 SMD for oral opioids, and ≥0.917 for intrathecal opioids) and statistically significant (P<0.001).
      Among patients treated with intrathecal opioids, an estimated 48.9% (95% CI, 35.7%–62.3%) of patients had at least a 30% reduction in pain and an estimated 40.0% (95% CI, 23.0%–59.8%) had at least a 50% reduction in pain by the longest follow-up time. In both bodies of evidence, longest follow up ranged from six months to a mean of 29 months. Findings from studies reporting at least a 30% reduction in pain were not substantially heterogenous (I2<0.001%) and no funnel plot asymmetry was detected using the trim-and-fill method. Heterogeneity (I2=69.8%) was detected among studies reporting at least a 50% reduction in pain, but could not be explained by duration of treatment time.

      Conclusions and Discussion

      Many patients in the included studies were so dissatisfied with adverse events or insufficient pain relief from opioids that they withdrew from the studies. For patients able to continue on opioids, evidence (albeit weak) suggests that their pain scores were lower than before therapy began and that this relief could be maintained long-term (≥6 months). However, data describing long-term safety and efficacy of opioids for CNCP are limited in terms of quantity and quality, precluding the formation of evidence-based conclusions supported by strong qualitative or stable quantitative evidence. An evidence base of low quality provides only weak evidence from which to draw qualitative conclusions and only low-stability evidence from which to draw quantitative conclusions. Some of the quantitative estimates were not robust upon sensitivity analyses, which means that an estimate of the size of a treatment effect cannot be accurately estimated with the currently available evidence. These quality ratings indicate that the evidence supporting our conclusions is highly subject to change, and that the likelihood is high that findings of future studies may overturn these conclusions. There is insufficient data to determine whether transdermal opioids maintain long-term pain relief. These findings are summarized in Table 3.
      Table 3Summary of Findings
      OutcomeMode of AdministrationKNI2Meta-analysis UnitFixed Effects Meta-analyses (95% CI)Funnel Plot Asymmetry?Random Effects Meta-analyses (95% CI)Quantitatively Robust to All Sensitivity Analyses?P-Value?
      Withdrawal due to adverse eventsOral61,40482.9%Event rate
      Denotes that this statistic could not be computed for that outcome because of the presence of heterogeneity.
      32.5% (26.1%–39.6%)Yes
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      Transdermal31,39998.2%Event rate17.5% (6.5% –39.0%)No
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      Intrathecal61540.00%Event rate6.3% (2.9%–13.1%)Yes6.3% (2.9%–13.1%)Yes
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      Withdrawal due to insufficient pain reliefOral71,50485.5%Event rate11.9% (7.8%–17.7%)Yes
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      Transdermal31,39948.9%Event rate5.8% (4.2%–7.3%)No5.8% (4.2%–7.3%)Yes
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      Intrathecal511063.3%Event rate10.5% (3.5%–27.4%)Yes
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      Pain
      Continuous outcomesOral316286.6%SMD on VAS (0–100), Likert (0–3; 0–4)1.99 (1.17–2.80)No<0.001
      Transdermal1
      In accordance with a priori protocols,14 evidence was not pooled when the evidence base was composed of fewer than three studies.
      Intrathecal612351.3%SMD on VAS (0–10; 0–100)1.33 (0.97–1.69)No<0.001
      ≥30% Pain reliefOral0
      Transdermal0
      Intrathecal3530.00%Event rate48.9% (35.7%–62.3%)No48.9% (35.7%–62.3%)Yes
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      ≥50% Pain reliefOral2
      In accordance with a priori protocols,14 evidence was not pooled when the evidence base was composed of fewer than three studies.
      Transdermal0
      Intrathecal510469.8%Event rate40.0% (23.0%–59.8%)Yes
      P-values not calculated for dichotomous outcomes, given absence of control groups.
      Abbreviations: K =studies, N=patients, CI=Confidence Interval, SMD=Standardized Mean Difference, VAS=Visual Analog Scale.
      a Denotes that this statistic could not be computed for that outcome because of the presence of heterogeneity.
      b P-values not calculated for dichotomous outcomes, given absence of control groups.
      c In accordance with a priori protocols,
      • Allan L.
      • Richarz U.
      • Simpson K.
      • Slappendel R.
      Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
      evidence was not pooled when the evidence base was composed of fewer than three studies.
      The generalizability of findings of these studies to “real-world” patients with chronic noncancer pain in general is unclear. Prescreening of patients in intrathecal studies for opioid responsiveness prior to commencement of treatment may limit the generalizability of the findings of these studies to patients who are not prescreened (neither the oral nor the transdermal studies had comparable prescreening requirements). In addition, the follow-up data on pain scores reported reflect outcomes for only a portion of the patients initially enrolled: many patients withdrew from studies that administered oral or transdermal opioids before they reported efficacy outcomes, and the majority of those studies did not report intent-to-treat outcome data. To investigate the importance of the possibility that the baseline characteristics of completers may differ in important ways from the baseline characteristics of the intent-to-treat population, we explored the effect of varying some input in sensitivity analyses. When baseline pain level standard deviations were changed, the summary effect size for oral opioids and intrathecal studies was not robust. When the assumed correlation coefficient was varied, the summary effect size of pain relief from intrathecal opioids was not robust either.
      Whether publication bias or other possible causes of missing studies (and, therefore, missing patients and outcomes data) influenced our results was not demonstrated by a funnel plot or trim-and-fill analysis. However, there may be a particular risk of publication bias in uncontrolled time-series studies, as journals may be less likely to accept studies of this design, investigators may be less likely to submit them for publication (given the lesser financial investment in conducting them), and no clinical trial registry for uncontrolled studies is currently in widespread use.
      Among outcomes for which qualitative conclusions were possible, substantial unexplained heterogeneity limited the number of quantitative conclusions we could make. Heterogeneity was encountered most frequently in the evidence bases for oral and transdermal opioid studies. Intrathecal evidence bases were generally not substantially heterogenous, perhaps due to the more rigorous prescreening for study enrollment and automation of dose administration. We postulated that at least some of the heterogeneity might be explained by differences in duration of treatment; however, meta-regression did not support this possibility for any outcome. Additional explorations of heterogeneity were not undertaken, given the limited number of available studies, and poor reporting. Whether the study was sponsored by drug or device manufacturers with a financial interest in the outcome would also be an interesting covariate; however, in many of these studies, financial sponsorship (or lack thereof) was not reported. Researchers whose studies are funded by parties without a conflict of interest in the findings should clearly state so in their articles. When more data become available, meaningful covariates for investigation of heterogeneity would include specific opioid administered, mode of administration, flexibility in dosing, use of adjuvant therapies, and etiology and baseline severity of pain. For the investigation of patient characteristics, publication of patient-level data would be even more desirable, because it would increase statistical power for meta-analysis and may be less influenced by study characteristics. In addition, the high attrition rates call into question whether the patients reporting pain relief outcomes are representative of the patient population originally enrolled in the study. We found no studies that attempted to identify prognostic factors for drop out. Studies designed to examine patient and treatment-related factors predicting long-term success with opioid therapy would be extremely useful.
      Opioid addiction is of great concern medically and societally. Fear that an individual with CNCP may develop dependence on the drug during long-term administration represents a potential barrier to treatment. However, the rate of observed signs of opioid addiction was extremely low (one case from a total of 2,042 enrollees) in the body of evidence considered in this review. This rate may be even lower if no addictive behaviors occurred among the 683 patients enrolled in the seven studies that did not mention addiction rates at all. At less than 1%, the addiction rate is comparable to the findings of Porter and Jick's oft-cited 1980 retrospective chart review of a general population of 11,882 hospitalized patients who were treated with narcotics for any duration.
      • Rainov N.G.
      • Heidecke V.
      • Burkert W.
      Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
      Opioid abuse was also rare (conservatively, 3/685). These rates do not support the contention that potential iatrogenic opioid addiction and abuse should limit therapy for well-selected patients.
      The low rates of addiction and abuse reported in these studies may only be generalizable to patients without a history of addictive/abusive behaviors. Twelve of the studies in this review screened out patients with a history of opioid or substance addiction or abuse, and two more screened for unspecified psychological contraindications (which may have included addictive/abusive history). Additional prospective studies are needed to address iatrogenic addiction/abuse in other patient populations. Given the importance of abuse and addiction issues, all studies on the efficacy and safety of opioids should prospectively collect data on abuse and addiction using validated diagnostic criteria.
      Given the complexity of definitively diagnosing opioid addiction (see Ballentyne
      • Ballantyne J.C.
      Opioids for chronic pain: taking stock.
      ) and in the interest of capturing the overall effect of opioid therapy on quality of life, we sought to analyze health-related quality-of-life outcomes in this review. However, insufficient data were available to enable quantitative analysis for any mode of administration, according to our protocols.

      Additional Thoughts and Discussion from the Pain Research Community

      The Mayday Fund (web site: http://www.painandhealth.org/mayday/mayday-home.html), a foundation sponsoring research on the treatment of pain, asked ECRI Institute to perform a review of efficacy and safety of opioids in addition to attitudinal and legal barriers to treatment with opioids. The Milbank Memorial Fund collaborated in that effort by inviting an international group of pain researchers from clinical, social, and legal disciplines to help formulate key questions for the systematic review. When the original work was completed, representatives from these organizations and additional pain researchers were asked to critique the review and to make recommendations for future research.
      Many ethical and practical difficulties are associated with long-term RCTs for CNCP. Allowing rescue opioids (with the amount and frequency of dosage as an outcome) in placebo or nonopioid groups may circumvent ethical problems of withholding analgesics from individuals with CNCP. While RCTs would be ideal, adequately powered and well-designed prospective cohort studies can provide very useful information.
      • Geborek P.
      • Crnkic M.
      • Petersson I.F.
      • Saxne T.
      South Swedish Arthritis Treatment Group. Etanercept, infliximab, and leflunomide in established rheumatoid arthritis: clinical experience using a structured follow up programme in southern Sweden.
      • Mamdani M.
      • Rochon P.A.
      • Juurlink D.N.
      • et al.
      Observational study of upper gastrointestinal haemorrhage in elderly patients given selective cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs.
      • MacDonald T.M.
      • Morant S.V.
      • Goldstein J.L.
      • Burke T.A.
      • Pettitt D.
      Channelling bias and the incidence of gastrointestinal haemorrhage in users of meloxicam, coxibs, and older, non-specific non-steroidal anti-inflammatory drugs.
      Ideally, a study protocol would specify uniform diagnostic and data collection criteria (e.g., pain etiology, drugs prescribed, dosing regimens), and would mimic clinical practice (e.g., drug combinations could be used, drug changes could be made, drug dosage could be titrated slowly and adjusted, and adverse effects could be aggressively managed). Reasons for discontinuing opioid therapy must be carefully documented. Data must also be collected on additional long-term effects of opioids that are currently not well understood, including androgen deficiency and changes in immune function. Practical/pragmatic RCTs that collect this information would be even more useful, but costs might be prohibitive.
      • Tunis S.R.
      • Stryer D.B.
      • Clancy C.M.
      Practical clinical trials: increasing the value of clinical research for decision making in clinical and health policy.
      Existing regulatory and clinical databases could be used to look for long-term dose-dependent efficacy and adverse effects. This could include data from the European Medicines Evaluation Agency (EMEA), Food and Drug Administration (FDA), Australia's Therapeutic Goods Administration, and other regulatory agencies. Studies on prescribing of opioids for noncancer pain performed using pharmacy and clinical data from the U.S. Veterans Healthcare Administration are examples of retrospective analyses.
      • Ytterberg S.R.
      • Mahowald M.L.
      • Woods S.R.
      Codeine and oxycodone use in patients with chronic rheumatic disease pain.
      • Reid M.C.
      • Engles-Horton L.L.
      • Weber M.B.
      • et al.
      Use of opioid medications for chronic noncancer pain syndromes in primary care.
      • Mahowald M.L.
      • Singh J.A.
      • Majeski P.
      Opioid use by patients in an orthopedics spine clinic.
      • Hermos J.A.
      • Young M.M.
      • Gagnon D.R.
      • Fiore L.D.
      Characterizations of long-term oxycodone/acetaminophen prescriptions in Veteran patients.
      Prospective data collection in such integrated systems could provide more reliable estimates of long-term efficacy and safety. Data from more broadly representative populations, however, would be important.
      There is agreement that the “best possible evidence” for therapeutic efficacy is provided by systematic reviews of high-quality RCTs. This systematic review has shown that such evidence is currently lacking for the efficacy and safety of long-term opioid therapy in chronic noncancer pain. As noted above, there are many practical difficulties in undertaking long-term RCTs in this patient population and there is increasing recognition in the evidence-based medicine community that other study designs (e.g., practical clinical trials
      • Tunis S.R.
      • Stryer D.B.
      • Clancy C.M.
      Practical clinical trials: increasing the value of clinical research for decision making in clinical and health policy.
      ) may be necessary for answering certain clinical questions. Although additional prospective controlled trials of long-term opioid treatment would provide valuable information, a prospective cohort study of suitable size has the potential to provide practical answers to key efficacy and safety questions.

      Acknowledgments

      The authors thank Vivian H. Coates, MBA, and Jeffrey Lerner, PhD, of ECRI Institute for overseeing this project from inception to completion, Eileen Erinoff of ERCI Institute for designing and conducting the systematic literature searches, and Clarisse Akafomo, MD, MPH, of ECRI for assistance in data acquisition. They also thank Philip J. Wiffen, MSc, MRPharmS, of the Cochrane Pain, Palliative and Supportive Care Review Group, Pain Research Unit, for contributions to the conception and design of the review and for reviewing earlier drafts of it, and Charles M. Turkelson, PhD for his contributions to the original project. They appreciate the members of the pain research, legal, social science, and regulatory community for their valuable insights during the earlier systematic review, and thank Dr. James Campbell, Dr. Howard Field, and Dr. Rollin Gallagher for a number of helpful comments on the scientific and clinical context for the review. The authors also wish to thank the following people who reviewed and provided comments on the original draft report: Steven I. Altchuler, PhD, MD, Daniel B. Carr, MD, Roger Chou, MD, MPH, Michael Clarke, DPhil, Wilson M. Compton, MD, MPE, Richard A. Denisco, MD, MPH, Kathleen M. Foley, MD, James Giglio, Geoff Gourlay, PhD, Marcia Meldrum, PhD, Robert Meyer, MD, Charles D. Ponte, PharmD, CDE, BCPS, Russell Portenoy, MD, Andrew S.C. Rice, MD, Linda Simoni-Wastila, PhD, Philip J. Wiffen, MSc, MRPharmS, and James Zacny, PhD, and those who participated in teleconferences on setting the research agenda: Philip J. Wiffen, MSc, MRPharmS, Chair, Mitchell Max, MD, Elon Eisenberg, MD, Ian Gilron, MD, MSc, FRCP (C), Geoff Gourlay, PhD, Art Lipman, PhD, Marcia Meldrum, PhD, Kathleen Foley, MD, Roger Goucke, FANZCA., FFPMANZCA, FAChPM, and Henry McQuay, MD. The full report of this review will be published in the Cochrane Library.

      References

        • Harstall C.
        • Ospina M.
        How prevalent is chronic pain?.
        Pain Clin Updat. 2003; 11: 1-4
        • Breivik H.
        • Collett B.
        • Ventafridda V.
        • Cohen R.
        • Gallacher D.
        Survey of chronic pain in Europe: prevalence, impact on daily life, and treatment.
        Eur J Pain. 2006; 10: 287-333
        • Gureje O.
        • Korff Von
        • Simon G.E.
        • Gater R.
        Persistent pain and well-being: a World Health Organization Study in Primary Care.
        JAMA. 1998; 280: 147-151
        • Clark J.D.
        Chronic pain prevalence and analgesic prescribing in a general medical population.
        J Pain Symptom Manage. 2002; 23: 131-137
        • Blyth F.M.
        • March L.M.
        • Brnabic A.J.
        • et al.
        Chronic pain in Australia: a prevalence study.
        Pain. 2001; 89: 127-134
        • Elliott A.M.
        • Smith B.H.
        • Penny K.I.
        • Smith W.C.
        • Chambers W.A.
        The epidemiology of chronic pain in the community.
        Lancet. 1999; 354: 1248-1252
        • Bowsher D.
        • Rigge M.
        • Sopp L.
        Prevalence of chronic pain in the British population: a telephone survey of 1037 households.
        Pain Clin. 1991; 4: 223-230
        • American Academy of Pain Medicine, American Pain Society
        The use of opioids for the treatment of chronic pain. A consensus statement from the American Academy of Pain Medicine and the American Pain Society.
        American Academy of Pain Medicine, American Pain Society, Glenview, IL1996
        • Jovey R.D.
        • Ennis J.
        • Gardner-Nix J.
        • et al.
        Use of opioid analgesics for the treatment of chronic noncancer pain—a consensus statement and guidelines from the Canadian Pain Society, 2002.
        Pain Res Manag. 2003; 8: 3A-28A
        • Kalso E.
        • Allan L.
        • Dellemijn P.L.
        • et al.
        Recommendations for using opioids in chronic non-cancer pain.
        Eur J Pain. 2003; 7: 381-386
        • Ballantyne J.C.
        • Mao J.
        Opioid therapy for chronic pain.
        N Engl J Med. 2003; 349: 1943-1953
        • Furlan A.D.
        • Sandoval J.A.
        • Mailis-Gagnon A.
        • Tunks E.
        Opioids for chronic noncancer pain: a meta-analysis of effectiveness and side effects.
        CMAJ. 2006; 174: 1589-1594
        • Eisenberg E.
        • McNicol E.
        • Carr D.B.
        Opioids for neuropathic pain.
        ([Review])The Cochrane Database of Systematic Reviews [internet]; Issue 3. John Wiley & Sons, Ltd., Hoboken, NJ2006https://doi.org/10.1002/14651858.CD006146 ([accessed 2006 Jul 26][Art. No. CD006146])
        • Allan L.
        • Richarz U.
        • Simpson K.
        • Slappendel R.
        Transdermal fentanyl versus sustained release oral morphine in strong-opioid naive patients with chronic low back pain.
        Spine. 2005; 30: 2484-2490
        • Eich E.
        • Reeves J.L.
        • Jaeger B.
        • Graff-Radford S.B.
        Memory for pain: relation between past and present pain intensity.
        Pain. 1985; 23: 375-380
        • Linton S.J.
        • Gotestam K.G.
        A clinical comparison of two pain scales: correlation, remembering chronic pain, and a measure of compliance.
        Pain. 1983; 17: 57-65
        • Agency for Healthcare Research and Quality (AHRQ)
        Systems to rate the strength of scientific evidence: Summary.
        Agency for Healthcare Research and Quality (AHRQ), Rockville, MD2002 March (11 p. (Evidence report/technology assessment; no. 47))
        • Egger M.
        • Juni P.
        • Bartlett C.
        • Holenstein F.
        • Sterne J.
        How important are comprehensive literature searches and the assessment of trial quality in systematic reviews? Empirical study.
        Health Technol Assess. 2003; 7: 1-76
        • Deeks J.J.
        • Dinnes J.
        • D'Amico R.
        • et al.
        Evaluating non-randomised intervention studies.
        Health Technol Assess. 2003; 7: 1-186
        • Treadwell J.T.
        • Tregear S.J.
        • Reston J.T.
        • Turkelson C.M.
        A system for rating the stability and strength of medical evidence.
        BMC Med Res Methodol. 2006; 6 (Available at): 52
      1. Sutton A.J. Abrams K.R. Jones D.R. Sheldon T. Song F. Methods for meta-analysis in medical research. John Wiley & Sons, 2001 January (274 p. (Wiley series in probability and mathematical statistics))
        • Hedges L.V.
        Fixed effects models.
        in: Cooper H. Hedges L.V. The handbook of research synthesis. Russell Sage Foundation, New York1994: 285-299
        • Raudenbush S.W.
        Random effects models.
        in: Cooper H. Hedges L.V. The handbook of research synthesis. Russell Sage Foundation, New York1994: 301-321
        • Shadish W.R.
        • Haddock C.K.
        Combining estimates of effect size.
        in: Cooper H. Hedges L.V. The handbook of research synthesis. Russell Sage Foundation, New York1994: 261-277
        • Hedges L.V.
        • Vevea J.L.
        Fixed- and random-effects models in meta-analysis.
        Psychol Methods. 1998; 3: 486-504
        • Greenhouse J.B.
        • Iyengar S.
        Sensitivity analysis and diagnostics.
        in: Cooper H. Hedges L.V. The handbook of research synthesis. Russell Sage Foundation, New York1994: 383-409
        • Higgins J.P.
        • Thompson S.G.
        • Deeks J.J.
        • Altman D.G.
        Measuring inconsistency in meta-analyses.
        BMJ. 2003; 327: 557-560
        • Thompson S.G.
        • Higgins J.P.
        How should meta-regression analyses be undertaken and interpreted?.
        Stat Med. 2002; 21: 1559-1573
        • Higgins J.P.
        • Thompson S.G.
        Controlling the risk of spurious findings from meta-regression.
        Stat Med. 2004; 23: 1663-1682
      2. Cooper H. Hedges L.V. The handbook of research synthesis. Russell Sage Foundation, New York1994
        • Lau J.
        • Schmid C.H.
        • Chalmers T.C.
        Cumulative meta-analysis of clinical trials builds evidence for exemplary medical care.
        J Clin Epidemiol. 1995; 48 (59–60): 45-57
        • Ioannidis J.P.
        • Contopoulos-Ioannidis D.G.
        • Lau J.
        Recursive cumulative meta-analysis: a diagnostic for the evolution of total randomized evidence from group and individual patient data.
        J Clin Epidemiol. 1999; 52: 281-291
        • Conti C.R.
        Clinical decision making using cumulative meta-analysis.
        Clin Cardiol. 1993; 16 ([Editorial]): 167-168
        • Thiessen Philbrook H.
        • Barrowman N.
        • Garg A.X.
        Imputing variance estimates do not alter the conclusions of a meta-analysis with continuous outcomes: a case study of changes in renal function after living kidney donation.
        J Clin Epidemiol. 2007; 60: 228-240
        • Duval S.
        • Tweedie R.
        Trim and fill: a simple funnel-plot-based method of testing and adjusting for publication bias in meta-analysis.
        Biometrics. 2000; 56: 455-463
        • Sutton A.J.
        • Duval S.J.
        • Tweedie R.L.
        • Abrams K.R.
        • Jones D.R.
        Empirical assessment of effect of publication bias on meta-analyses.
        BMJ. 2000; 320: 1574-1577
        • Duval S.J.
        • Tweedie R.L.
        A non-parametric ‘trim and fill’ method of assessing publication bias in meta-analysis.
        J Am Stat Assoc. 2000; 95: 89-98
        • Milligan K.
        • Lanteri-Minet M.
        • Borchert K.
        • et al.
        Evaluation of long-term efficacy and safety of transdermal fentanyl in the treatment of chronic noncancer pain.
        J Pain. 2001; 2: 197-204
        • Mystakidou K.
        • Parpa E.
        • Tsilika E.
        • et al.
        Long-term management of noncancer pain with transdermal therapeutic system-fentanyl.
        J Pain. 2003; 4: 298-306
        • Hassenbusch S.J.
        • Stanton-Hicks M.
        • Covington E.C.
        • Walsh J.G.
        • Guthrey D.S.
        Long-term intraspinal infusions of opioids in the treatment of neuropathic pain.
        J Pain Symptom Manage. 1995; 10: 527-543
        • Harati Y.
        • Gooch C.
        • Swenson M.
        • et al.
        Maintenance of the long-term effectiveness of tramadol in treatment of the pain of diabetic neuropathy.
        J Diabetes Complications. 2000; 14: 65-70
        • Caldwell J.R.
        • Rapoport R.J.
        • Davis J.C.
        • et al.
        Efficacy and safety of a once-daily morphine formulation in chronic, moderate-to-severe osteoarthritis pain: results from a randomized, placebo-controlled, double-blind trial and an open-label extension trial.
        J Pain Symptom Manage. 2002; 23: 278-291
        • McIlwain H.
        • Ahdieh H.
        Safety, tolerability, and effectiveness of oxymorphone extended release for moderate to severe osteoarthritis pain: a one-year study.
        Am J Ther. 2005; 12: 106-112
        • Peloso P.M.
        • Bellamy N.
        • Bensen W.
        • et al.
        Double blind randomized placebo control trial of controlled release codeine in the treatment of osteoarthritis of the hip or knee.
        J Rheumatol. 2000; 27: 764-771
        • Anderson V.C.
        • Burchiel K.J.
        A prospective study of long-term intrathecal morphine in the management of chronic nonmalignant pain.
        Neurosurgery. 1999; 44 (discussion 300–301): 289-300
        • Anderson V.C.
        • Burchiel K.J.
        • Cooke B.
        A prospective, randomized trial of intrathecal injection vs. epidural infusion in the selection of patients for continuous intrathecal opioid therapy.
        Neuromodulation. 2003; 6: 142-152
        • Angel I.F.
        • Gould Jr., H.J.
        • Carey M.E.
        Intrathecal morphine pump as a treatment option in chronic pain of nonmalignant origin.
        Surg Neurol. 1998; 49 (discussion 98–99): 92-98
        • Fredheim O.M.
        • Kaasa S.
        • Dale O.
        • et al.
        Opioid switching from oral slow release morphine to oral methadone may improve pain control in chronic non-malignant pain: a nine-month follow-up study.
        Palliat Med. 2006; 20: 35-41
        • Kanoff R.B.
        Intraspinal delivery of opiates by an implantable, programmable pump in patients with chronic, intractable pain of nonmalignant origin.
        J Am Osteopath Assoc. 1994; 94: 487-493
        • Mironer Y.E.
        • Tollison C.D.
        Methadone in the intrathecal treatment of chronic nonmalignant pain resistant to other neuroaxial agents: the first experience.
        Neuromodulation. 2001; 4: 25-31
        • Rainov N.G.
        • Heidecke V.
        • Burkert W.
        Long-term intrathecal infusion of drug combinations for chronic back and leg pain.
        J Pain Symptom Manage. 2001; 22: 862-871
        • Tutak U.
        • Doleys D.M.
        Intrathecal infusion systems for treatment of chronic low back and leg pain of noncancer origin.
        South Med J. 1996; 89: 295-300
        • Zenz M.
        • Strumpf M.
        • Tryba M.
        Long-term oral opioid therapy in patients with chronic nonmalignant pain.
        J Pain Symptom Manage. 1992; 7: 69-77
        • Roth S.H.
        • Fleischmann R.M.
        • Burch F.X.
        • et al.
        Around-the-clock, controlled-release oxycodone therapy for osteoarthritis-related pain: placebo-controlled trial and long-term evaluation.
        Arch Intern Med. 2000; 160: 853-860
        • Matsumoto A.K.
        • Babul N.
        • Ahdieh H.
        Oxymorphone extended-release tablets relieve moderate to severe pain and improve physical function in osteoarthritis: results of a randomized, double-blind, placebo- and active-controlled phase III trial.
        Pain Med. 2005; 6: 357-366
        • Ballantyne J.C.
        Opioids for chronic pain: taking stock.
        Pain. 2006; 125: 3-4
        • Geborek P.
        • Crnkic M.
        • Petersson I.F.
        • Saxne T.
        South Swedish Arthritis Treatment Group. Etanercept, infliximab, and leflunomide in established rheumatoid arthritis: clinical experience using a structured follow up programme in southern Sweden.
        Ann Rheum Dis. 2002; 61: 793-798
        • Mamdani M.
        • Rochon P.A.
        • Juurlink D.N.
        • et al.
        Observational study of upper gastrointestinal haemorrhage in elderly patients given selective cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs.
        BMJ. 2002; 325: 624
        • MacDonald T.M.
        • Morant S.V.
        • Goldstein J.L.
        • Burke T.A.
        • Pettitt D.
        Channelling bias and the incidence of gastrointestinal haemorrhage in users of meloxicam, coxibs, and older, non-specific non-steroidal anti-inflammatory drugs.
        Gut. 2003; 52: 1265-1270
        • Tunis S.R.
        • Stryer D.B.
        • Clancy C.M.
        Practical clinical trials: increasing the value of clinical research for decision making in clinical and health policy.
        JAMA. 2003; 290: 1624-1632
        • Ytterberg S.R.
        • Mahowald M.L.
        • Woods S.R.
        Codeine and oxycodone use in patients with chronic rheumatic disease pain.
        Arthritis Rheum. 1998; 41: 1603-1612
        • Reid M.C.
        • Engles-Horton L.L.
        • Weber M.B.
        • et al.
        Use of opioid medications for chronic noncancer pain syndromes in primary care.
        J Gen Intern Med. 2002; 17: 173-179
        • Mahowald M.L.
        • Singh J.A.
        • Majeski P.
        Opioid use by patients in an orthopedics spine clinic.
        Arthritis Rheum. 2005; 52: 312-321
        • Hermos J.A.
        • Young M.M.
        • Gagnon D.R.
        • Fiore L.D.
        Characterizations of long-term oxycodone/acetaminophen prescriptions in Veteran patients.
        Arch Intern Med. 2004; 164: 2361-2366