Criteria for considering studies for this review
Types of studies
Randomised controlled trials, with parallel-group or cross-over design., comparing oxycodone (any formulation and any route of administration) with placebo or an active drug (including oxycodone) for cancer background pain. We will not examine studies on breakthrough pain.
Types of participants
Adults (aged ≥ 18 years) with cancer pain.
Types of interventions
Oxycodone (any dose, formulation and route of administration) versus oxycodone (any dose, formulation and route of administration).
Oxycodone (any dose, formulation and route of administration) versus other active drug (any dose, formulation and route of administration).
Oxycodone (any dose, formulation and route of administration) versus placebo.
Types of outcome measures
Pain intensity and pain relief.
Both of these outcomes have to be patient-reported and can be reported in any transparent manner (e.g., by using numerical or verbal rating scales). We will not consider these outcomes reported by physicians, nurses or carers. If possible, we will distinguish between nociceptive and neuropathic pain.
Side effects or adverse events (e.g., constipation, nausea, vomiting, drowsiness, confusion, respiratory depression), quality of life, and patient preference.
We will consider all of these outcomes as they are reported in the included studies.
Search methods for identification of studies
We will not apply language, date or publication status (published in full, published as abstract, unpublished) restrictions to the search.
We will identify relevant trials by searching the following databases:
The Cochrane Central Register of Controlled Trials (CENTRAL)
MEDLINE (OVID) (1946 to date)
MEDLINE in process
EMBASE (OVID) (1947 to date)
Science Citation Index (Web of Science) (1899 to date)
Conference Proceedings Citation Index - Science (Web of Science) (1990 to date)
BIOSIS (Web of Science) (1926 to date)
PsycINFO (OVID) (1806 to date)
We will apply the Cochrane highly sensitive search strategy for identifying randomised control trials to this search (Lefebvre 2011). The search strategy for MEDLINE is in Appendix 1 and will be modified for other databases using the appropriate syntax and controlled vocabulary.
Searching other resources
We will check the bibliographic references of relevant identified studies in order to find additional trials not identified by the electronic searches. We will also search Clinicaltrials.gov and metaRegister of Controlled Trials (mRCT) as complementary sources for related studies. We will contact authors of the included studies to ask if they know of any other relevant studies.
Data collection and analysis
Selection of studies
Two of the review authors (MSH, MIB) will assess the titles and abstracts of all the studies identified by the search for potential inclusion. We will independently consider the full records of all potentially relevant studies for inclusion by applying the selection criteria outlined in the Criteria for considering studies for this review section. We will resolve potential disagreements by discussion. We will not restrict the inclusion criteria by date, language or publication status (published in full, published as abstract, unpublished).
Data extraction and management
Using a standardised data extraction form, two authors (MSH, MIB) will extract data pertaining to study design, participant detail (including age, cancer characteristics, previous analgesic medication and setting), interventions (including details about titration), and outcomes. We will resolve potential disagreements by discussion. If there are studies for which only a subgroup of the participants meet the inclusion criteria for the current review, we will only extract data on this subgroup provided randomisation will not be broken.
Assessment of risk of bias in included studies
Two of the authors (MSH, MIB) will independently assess the methodological quality of each of the included studies by using the 'Risk of bias' assessment method outlined in the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2011). For each study we will assess the risk of bias for the following domains: Selection bias (study level; 2 items; random sequence generation, allocation concealment), performance bias (outcome level; 2 items; blinding of patients, blinding of treating personnel), detection bias (outcome level; 1 item; blinding of outcome assessment), attrition bias (outcome level; 1 item; incomplete outcome data), and reporting bias (study level; 1 item; selective reporting). We will also include an item that assesses the adequacy of titration. Each of the 'Risk of bias' items requires a 'low risk', 'high risk' or 'unclear risk' response. We will also document the reasons for each response in accordance with Higgins 2011. We will resolve potential disagreements between the 'Risk of bias' ratings through discussion.
Measures of treatment effect
For continuous outcomes we will extract the means and standard deviations and we will use these to estimate the mean difference between the treatments along with the 95% confidence interval (CI), if the outcome is measured on the same scale in the studies. Where the outcome is measured on different scales, we will report the standardised mean difference with 95% CIs instead. For dichotomous outcomes we will extract event rates and calculate risk ratios.
Unit of analysis issues
The patient will be the unit of analysis, but if the data reported in any included cross-over trials cannot be otherwise incorporated into the analyses (see Dealing with missing data), we will include them as if the design had been parallel group. Higgins 2011 (in chapter 16) points out that this approach, while giving rise to unit-of analysis error, is nevertheless conservative as it results in an under-weighting of the data. If we include cross-over trial data in this manner we will perform sensitivity analyses assessing the impact of this strategy.
Dealing with missing data
In cases where data are missing, we will contact the authors to request the missing data. Missing data imputation will be limited to the imputation of missing standard deviations if enough information is available from the studies to calculate the standard deviation according to the methods outlined by Higgins 2011. We will record the drop-out/missing data rates in the 'Risk of bias' tables under the items on attrition bias, and we will address the potential effect of the missing data on the results in sensitivity analyses and in the Discussion section of the review. In all cases we will aim to perform intention-to-treat analyses.
Assessment of heterogeneity
We will assess heterogeneity by using the I2 statistic. We will consider I2 values above 50% to represent substantial heterogeneity in line with Higgins 2011 and we will assess potential sources of heterogeneity through subgroup analyses as outlined in Subgroup analysis and investigation of heterogeneity.
Assessment of reporting biases
In addition to implementing the comprehensive search strategy outlined in the section Search methods for identification of studies, the risk of outcome reporting bias will be illustrated in the 'Risk of bias' summary figures that we will construct for each study and each type of assessed bias.
We will enter the data extracted from the included studies into Review Manager (RevMan 2012) which will be used for data synthesis. We will analyse continuous outcomes using the generic inverse variance method, and dichotomous outcomes using the Mantel-Haenszel method in accordance with Higgins 2011. If I2 is above 50% we will use a random-effects model and consider not reporting a summary estimate of the data (depending on the subgroup analyses; see also the section Subgroup analysis and investigation of heterogeneity). Otherwise we will use a fixed-effect model for the meta-analyses. If it is not feasible to meta-analyse the data from the included studies, we will summarise the data narratively and in tables, if sensible.
Subgroup analysis and investigation of heterogeneity
Different aspects of the trials are likely to contribute heterogeneity to the proposed main analyses. If there are sufficient data, we will therefore perform subgroup analyses based on doses, titration, formulations (e.g., immediate-release, sustained-release), routes of administration (e.g. oral, rectal), length of the trials, and populations (e.g. adults, opioid-naive patients).
If sufficient data are available, we will examine the robustness of the meta-analyses by conducting sensitivity analyses using different components of the 'Risk of bias' assessment, particularly those relating to whether allocation concealment and blinding were adequate. We will conduct further sensitivity analyses to examine the impact of missing data on the results if a large proportion of the studies are at an 'unknown' or 'high risk' of attrition bias, and finally, sensitivity analyses will examine whether publication status and trial size influence the results.