McMaster University

Michael G. DeGroote
National Pain Centre

Scope of Search

Canadian Guideline for Safe and Effective Use of Opioids for Chronic Non-Cancer Pain

10. Literature Search Methods

Development of this Canadian Guideline relied on the 2006 meta-analysis by Furlan et al. "Opioids for chronic non-cancer pain: a meta-analysis of effectiveness and side effects" (Furlan 2006). In addition, three new literature searches were completed:

  • Search One: Search for randomized controlled trials (RCT) published since May 2006 to update the Furlan meta-analysis.
  • Search Two: Search for additional literature (multiple designs) that answered questions about the treatment of CNCP with opioids and managing the patient with problematic opioid use.
  • Search Three: Search for additional literature (multiple designs) that answered questions about long-term outcomes of opioid use.

10.1 Description of Literature Search One

For details of the original Furlan meta-analysis search (Furlan 2006), see http://www.cmaj.ca/cgi/data/174/11/1589/DC1/1 and http://www.cmaj.ca/cgi/data/174/11/1589/DC1/10.

The following bibliographic data sources were used to update the review to July 2009:

  • Cochrane Central Register of Controlled Trials (CENTRAL) 2009
  • MEDLINE (OVID) from 2005 to July 2009 (same strategy as the 2006 review)
  • EMBASE from 2005 to July 2009 (same strategy as the 2006 review)
  • Reference lists of retrieved articles
  • Articles forwarded by the National Advisory Panel

Search strategies for MEDLINE and EMBASE are available (see Appendix A-4 Literature Search Strategies). A research librarian ran the electronic searches and coordinated the data entry into Reference Manager® 11, removing all duplicates.

10.1.1 Relevance Screening for Search One

Three CPSO research associates independently reviewed the titles and abstracts using the following criteria: 1) not a letter, editorial or short commentary (usually less than three pages in length); 2) focus of the article is not dealing with surgical pain, 3) article is not dealing with cancer pain, 4) population studied had chronic non-cancer pain, and 5) focus is on opioids. Studies that passed the relevance screen were forwarded to the Research Group for inclusion/exclusion criteria screening.

10.1.2 Inclusion/Exclusion Screening for Search One

Text of full articles was obtained for studies that passed the relevance screening. Two Research Group members independently reviewed these studies and applied inclusion/exclusion criteria as follows:

  • Study characteristics: Included RCTs published in English, French, Portuguese, or Spanish (languages that could be read by Research Group members). Excluded studies published only as abstracts.
  • Study population: Included adults (>18 years) with CNCP (defined as pain that persists for more than six months) including neuropathic pain, osteoarthritis, rheumatoid arthritis, fibromyalgia, and back and musculoskeletal pain. Excluded migraines, dental pain, ischemic pain due to vascular disease and abdominal pains (e.g., chronic pancreatitis, kidney stones) because these conditions are usually classified as CNCP.
  • Types of intervention: Included any opioid administered by oral, transdermal, transmucosal or rectal route for seven days or more. Opioids were classified as weak (propoxyphene, codeine, tramadol, hydrocodone) or strong (oxycodone, morphine, fentanyl, hydromorphone or buprenorphine). Excluded methadone.
  • Types of comparison group: Included placebo or other analgesics. Excluded comparisons of different opioids.
  • Outcomes: Quantifying pain (intensity or relief), function, and side effects.

For Search One, two reviewers reviewed selected titles, abstracts, and full texts and determined the articles for inclusion. If consensus could not be achieved, a third reviewer was consulted. On some occasions, authors of the randomized trials were contacted to obtain more details that were not reported in the publication.

10.1.3 Methodological Quality Screen for Search One

The same two Research Group members completed an independent appraisal of methodological quality on studies admitted after inclusion/exclusion screening. Where needed, they reached consensus through discussion. Reviewers were not blinded with respect to authors, institution and journal because they were familiar with the literature. In cases of disagreement, a third reviewer was consulted. Each study was scored from 0 to 5 with the instrument developed by Jadad and colleagues (Jadad 1996). The instrument includes three questions about randomization methods, double-blinding, and number of withdrawals. Studies scoring 3, 4, or 5 were considered to be of high quality; scoring 0, 1, or 2, of low quality. Study scores were recorded in a Microsoft Excel® spreadsheet (see Appendix B-13, Part B).

10.1.4 Data Extraction and Synthesis for Search One

Research Group members extracted the data from the high quality studies using Microsoft Excel®. Meta-analyses and meta-regression were conducted using Comprehensive Meta-Analysis© software, with calculations of effect sizes for pain relief and functional outcomes.

Effect Size: Cohen’s three levels (Cohen 1988) were used and adapted to a scale developed by the Cochrane Back Review Group (Furlan 2009):

  • Small = ES <0.5 = Mean difference less than 10% of the scale (e.g., <10mm on a 100mm visual analog scale)
  • Medium = ES from 0.5 to <0.8 = Mean difference 10 to 20% of the scale
  • Large = ES ≥0.8 = Mean difference >20% of the scale

For side effects, all meta-analyses were done using RevMan 52 using risk differences. Statistical heterogeneity was tested by Q test (chi-square) reported as I2 (higher values indicate higher heterogeneity).

All meta-analyses were conducted using a random effects model. Sub-groups were decided a priori to assess the variations in effect sizes. Clinical significance of side effects was considered when the incidence was 10% or higher in the opioid or reference group.

2 Review Manager (RevMan) [Computer program]. Version 5.0. Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2008.

10.2 Description of Literature Search Two and Search Three

Search Two was conducted to find articles that could be useful in drafting the recommendations on the treatment of CNCP with opioids and managing the patient with problematic opioid use. Search Three was conducted to understand the effects of prolonged opioid use. These searches were not limited to RCTs. (See Appendix A-5 Flowchart of Literature Review Process and Appendix A-4: Literature Search Strategies.)

The following bibliographic data sources were used:

  • Cochrane Central Register of Controlled Trials (CENTRAL) 2009
  • MEDLINE (OVID) from 1950 to July 2009
  • EMBASE from 1982 to July 2009
  • Reference lists of retrieved articles
  • Articles forwarded by the National Advisory Panel

10.2.1 Relevance Screen for Search Two and Search Three

A CPSO research associate independently reviewed the titles and abstracts using the following criteria: 1) not a letter, editorial or short commentary (usually less than three pages in length), 2) population studied has chronic non-cancer pain, 3) focus on opioids, and 4) focus on addiction.

10.2.2 Inclusion/Exclusion Screen for Search Two and Search Three

From the titles and abstracts that passed the relevance screen, text of full articles was obtained, and two out of four Research Group members applied inclusion criteria:

  1. Study characteristics: Included any study design with primary data collection, conducted in humans, with no language restriction. Studies could be experimental (e.g., clinical trials), observational (e.g., cohort, case-control, cross-sectional) or descriptive (e.g., before-and-after, case series, case reports). Studies published in a language other than English were judged for inclusion/exclusion, based on the English abstract.
  2. Study population: Included adults (>18 years) with CNCP (defined as pain that persists for more than six months) including neuropathic pain, osteoarthritis, rheumatoid arthritis, fibromyalgia, and back and musculoskeletal pain. Excluded acute pain, post-surgical pain, or experimental pain in healthy volunteers. In some circumstances, a study in a population with cancer pain could be included if information could be extrapolated to non-cancer pain.
  3. Types of intervention: Included any opioid administered by oral, transdermal, transmucosal or rectal route for pain for seven days or more. Studies of methadone were included.
  4. Useful Topics: Included topics deemed to be of value in drafting the recommendations on the treatment of CNCP with opioids and managing the patient with problematic opioid use:
    • Dose of opioids to achieve maximum benefits with minimum adverse events
    • Urine drug screening
    • Initiation, titration and tapering of opioids
    • Assessments and monitoring during treatment with opioids
    • Frequency of follow-up
    • Identification of patients at risk for medical complications, overdose, misuse or addiction
    • Recommendations for practice regarding screening, management, follow-up
    • Approaches to dealing with conflicts with patients
    • Treating chronic pain patients in acute care settings
    • Mechanisms to prevent prescription fraud
    • Use of opioids and driving
    • Identifying patients at risk of opioid addiction
    • Managing an opioid addicted patient with chronic pain
    • Tapering and stopping opioids or other drugs, e.g., benzodiazepines
    • Dealing with challenging or threatening patients
    • Long-term outcomes of opioid use

    For Searches Two and Three, four reviewers worked in pairs to select articles for inclusion. When in doubt, a third reviewer from the other pair was consulted.

10.2.3 Additional Strategies for Search Two and Search Three

All included and excluded studies from Search One were also evaluated by two reviewers against the list of useful topics developed for inclusion of studies in the Searches Two and Three.

10.2.4. Methodological Quality Screen for Searches Two and Three

Observational studies were not assessed for methodological quality due to lack of resources to fund experts in epidemiological methods necessary to complete the more complex and subjective review required.

10.3 Using Extracted Evidence to Develop Recommendations for Practice

10.3.1. Recommendation Development Process

The Research Group provided methodological and clinical expertise in the area of chronic pain and addiction medicine . They summarized evidence from the studies and drafted 49 initial recommendations that each included a discussion and related evidence. An iterative course of action ensued, using a Modified Delphi technique with the National Advisory Panel (NAP), to produce final recommendations. NAP member identities were blind to the Research Group and each other until the last round of review.

NAP received material via email and responded using an on-line survey tool to rate their opinion on relevance, feasibility, clarity, and their degree of agreement with each recommendation. They also provided open-ended narrative comments.

Consensus was defined as 80% of NAP members supporting a recommendation. Recommendations that did not receive this level of consensus were revised using feedback provided by NAP and re-rated in the next round. With each round of review, each NAP member received a complete transcript of all written comments made by NAP in the previous round.

While participation rates declined as the Modified Delphi progressed, the portion of NAP members involved remained high throughout, as summarized in Table A-10.3.1. A drop in the last two rounds could have been due to Panel fatigue, or related to the H1N1 pandemic occurring in Canada at the time. Consensus on recommendations resulted after four rounds of electronic review and rating, culminating with a final telephone and web-assisted meeting.

Table A-10.3.1 National Advisory Panel Participation in Modified Delphi Process

Round Number of recommendations under review Panelists participating
1 49 84%
2 20 80%
3 4 65%
4 2 60%

10.3.2 Recommendation Grading

The evidence-grading system was adapted from the Canadian Task Force on Preventive Health Care (CTFPHC) (Woolf 1990). (See Table A-10.3.2.) A single recommendation statement can be supported by one, two, or three different grades of evidence.

Each recommendation topic includes a key word, recommendation statement, discussion, and evidence summary. References may be provided in both the discussion and evidence summary. There are two types of references used: those that 1) provide direct or indirect support for the recommendation statement and 2) provide contextual information.

If a reference supported directly, the recommendation statement was graded consistent with the study design of that reference, i.e., "A" or "B." (See Table A-10.3.2.)

If a reference supported indirectly, the recommendation statement was graded to reflect the primary source driving the recommendation.

  • Example 1: a RCT informed the recommendation but the recommendation is graded "B" or "C" (rather than "A") — this is because the recommendation statement is not directly extracted from the main hypothesis of the RCT.
  • Example 2: references are graded "B" in the evidence summary, but the recommendation statement is graded "C" — this is because expert opinion from NAP was the predominant driver of the recommendation statement, even though some of the recommendation’s concepts were backed by the studies mentioned in the evidence summary.
  • Example 3: a reference conflicts with the recommendation, and the recommendation statement is graded "C" — this reflects NAP expert opinion assessing the evidence as weak or not generalizable.

Table A-10.3.2 Recommendation Grading

CTFPHC Evidence Grading System (Woolf 1990) Canadian Guideline Recommendation Grading
I – Evidence from RCTs.

Grade A: Recommendations are supported by evidence from RCT(s).

II – 1 Evidence from controlled trial(s) without randomization.

II – 2 Evidence from cohort or case-control analytic studies, preferably from more than one centre or research group.

II – 3 Evidence from comparisons between times or places with or without the intervention; dramatic results from uncontrolled studies could be included here.

Grade B: Recommendations are supported by:

  • Evidence from controlled trial(s) without randomization, or,
  • Evidence from cohort or case-control analytic studies, preferably from more than one centre or research group, or
  • Evidence from comparisons between times or places with or without the intervention; dramatic results in uncontrolled experiments could be included here.

III – Opinions of respected authorities, based on clinical experience; descriptive studies or reports of expert committees.

Grade C: Recommendations are supported by consensus opinion of the National Advisory Panel.