Abstract
AIMS
To investigate the use of an enriched study design on the estimates of treatment effect in analgesic trials for chronic musculoskeletal pain.
Methods
Database searches were conducted from 2004 to 2014. We included randomized placebo‐controlled trials evaluating pain medications for chronic musculoskeletal pain. Methodological quality was assessed using the PEDro scale. The estimates of treatment effect on pain and adverse events were compared between enriched and nonenriched designs. Metaregression was used to assess the association between the effect size estimate and the study design controlling for analgesic dose and methodological quality.
Results
We included 108 trials, of which 99 were included in the meta‐analysis (n = 44 171). There were no overall differences in effect sizes between enriched and nonenriched designs for pain intensity. There was a significant difference for a reduction in any adverse events favouring enriched designs for opioids, but not for other analgesics or the outcome serious adverse events. There was an association between effect size and methodological quality, with failure to blind the outcome assessor and failure to use intention‐to‐treat analysis being associated with larger effect sizes.
Conclusions
There is no evidence that the use of an enriched study design changes the treatment effect size estimate for pain. There is some evidence that clinical trials that employ enriched designs report a reduced risk of adverse events in trials for chronic musculoskeletal pain, but it is unclear whether enriched designs influence estimates of serious adverse events. Features of trial design and study quality were associated with treatment effect estimates.
Keywords: analgesics, chronic pain, methodology, opioids
What is Already Known about this Subject
Enriched enrolment designs aim to increase the proportion of treatment responders to a treatment and to decrease the number of participants who do not tolerate the treatment.
It is unknown whether enriched designs can influence treatment effect sizes or reduce adverse events.
What this Study Adds
Enriched enrolment designs do not influence the effect estimates for pain intensity in analgesic trials of chronic pain.
There is some evidence that enriched enrolment designs in chronic pain yield a reduced risk of adverse events.
Methodological features such as assessor blinding and intention‐to‐treat analysis were associated with effect estimates.
Introduction
Randomized controlled trials have a key role in determining the effects of interventions in healthcare. However, intervention effects are often relatively small and difficult to distinguish from possible study biases 1. One option to address this issue has been the practice of restricting enrolment in clinical trials to participants more likely to respond to the intervention by using an enriched study design. Enriched enrolment studies are a type of randomized controlled trial that aims to increase the proportion of so‐called treatment responders to a study medicine and to decrease the number of participants who do not tolerate the study treatment 2.
Various enrichment study designs have been used in past research. In the classical approach, participants receive the study drug for a period of time prior to randomization (called an open‐label or run in phase) aiming to identify patients who respond to the drug. Then, those patients who demonstrate benefit (improvement in study endpoints) from the drug and can tolerate it (the so‐called enriched population) are included in the trial and randomized to either continue in the treatment group or to receive the alternate intervention (usually a placebo) 3, 4, 5. A variation on this approach is to perform a preliminary trial instead of an open label phase to identify responders to the drug; and then enrol only those responders in a second, placebo‐controlled trial 2. Another alternative is a flare study design, in this case only those patients who exhibit a flare of their symptoms when current treatment is ceased (in an open‐label phase) are selected to enter the trial 2, 5.
A proposed benefit of selecting the responders to the study treatment is that they are more likely to continue the treatment in actual clinical practice and so proponents of enriched study designs argue that the trial results would be more clinically relevant 5, 6. This study design can also have the potential to increase the response rate and reduce drop‐outs related to drug intolerance or adverse effects 2, 7, 8, 9. However, opponents of an enriched study design point out that the generalizability of the results may be problematic. The efficacy in a subpopulation of responders may not predict the treatment response in the general population, since the general population consists of a mix of participants, including both responders and nonresponders to the study drug 4, 6. Additionally, valid information to identify who will respond (or not) to the drug a priori is often lacking.
Previous systematic reviews of enriched study designs have demonstrated conflicting results on whether the enriched design actually does yield larger estimates of treatment effect 3, 10. These reviews, however, were restricted to one type of medicine and/or specific conditions [opioids for chronic pain and nonsteroidal anti‐inflammatories (NSAIDs) for osteoarthritis, respectively]. A broader review is needed to understand the influence of this design on the effect estimates of pain medication for chronic musculoskeletal pain. Therefore, the aim of this study is to compare the impact of enriched and nonenriched study designs on the effect size estimates of various analgesics in trials of chronic musculoskeletal pain. This study will include any type of analgesic for nonspecific neck or back pain or osteoarthritis. We hypothesize that trials that use an enriched study design will report significantly larger treatment effect sizes and fewer adverse events when compared to clinical trials that do not use an enriched study design.
Methods
The protocol for this review was registered prospectively on PROSPERO (CRD42014009988). Because the treatment effect size could be influenced by the health condition, type of medicine and the treatment outcomes studied in the trial, we restricted the inclusion of trials to have a more homogeneous set of trials. As effects of medicines may reduce over time we conducted separate analyses for short, intermediate and long‐term results. As there is evidence that trial methodological quality influences estimates of treatment effect we included methodological quality as a covariate in the analyses.
Data sources and searches
We searched for reports of randomized controlled trials indexed in MEDLINE, EMBASE and the Cochrane Central Register of Controlled Trials (CENTRAL) and published in a 10‐year period (2004 to July 2014). We only included full text papers written in English, Portuguese, Dutch and Spanish as the authors spoke these languages. The search strategy is described in Appendix S1. We also performed citation tracking of the reference lists of relevant systematic reviews identified by our search and also of the included trials.
Trial selection
Reflecting the aim of this study (i.e. to compare enriched and nonenriched study designs), we included only randomized controlled trials evaluating analgesics in the treatment of chronic musculoskeletal pain compared to placebo. Trials that used quasirandom allocation procedures (e.g. alternation or by medical record number) were not included in this review.
The inclusion criteria were participants (no age limit) with chronic (duration of symptoms 3 or more months) nonspecific neck pain or nonspecific low back pain or osteoarthritis of the hip or knee. Nonspecific neck pain or back pain refers to spinal pain without any known specific cause or pathology (e.g. fracture, cancer and inflammatory diseases) 11. For osteoarthritis, trials using either a clinical (e.g. based on clinical American College of Rheumatology criteria) or radiological (e.g. x‐ray, medical resonance imaging) diagnosis were included, with no distinction between them.
We considered any analgesic (e.g. paracetamol, opioid analgesics and NSAIDs) used for treating musculoskeletal pain. However, we categorized the type of drug into opioids or other analgesics (i.e. NSAIDs, paracetamol, combinations of analgesics, and complementary and alternative medicines) for the analyses. The opioid analgesic dosages were converted to morphine equivalent units 12, as used in a previous study 7. We anticipated using the Hill equation to characterize the dose–response relationship for the study medicines 13; however, as the analgesics did not differ from the recommended daily dosage, the correction was not necessary. No restriction was placed on the setting of care or context of the included trials.
The primary outcome was pain intensity measured by any reliable and valid self‐report continuous outcome measure (such as a visual analogue scale or a numerical rating scale) at follow‐up. Secondary outcomes were the frequency of any adverse events and serious adverse events.
We included any form of enriched study design (e.g. responders and flare designs). Complete and partial enriched enrolment trials were considered. Complete enriched enrolment would occur in two circumstances: (i) inclusion restricted to participants responding to the test drug or a closely related drug with similar mechanism of action, or (ii) exclusion of participants who do not respond to the test drug, a closely related drug (and when all participants had been exposed to the drug), or those who do not tolerate the drug (e.g. presence of side effects). Partial enriched enrolment was defined as the exclusion from the trial of any previous nonresponders to the drug or a similar drug, but where not all participants were known to have been exposed to the drug. All the other forms of enrolment were defined as nonenriched enrolment. In other words, when no statement of inclusion or exclusion of participants could be interpreted as enriching the sample to responders to the drug treatment 2.
Data extraction and quality assessment
We extracted the following data from each of the eligible trials using a standardized data extraction form: bibliometric data (authors, title, source, year of publication, language); trial characteristics (design, sample size in the open‐label or run in phase, number of participants in the enriched/flare phase, description of the sample, country, recruitment modality, funding); characteristics of the participants (sex, age, region of pain, duration of symptoms, severity of the condition at baseline); description of the drugs, including route of administration, dose form (tablet, extended release), dose regimen (mg h–1 day–1), the duration of treatment, classification of medicine by World Health Organisation Anatomical Therapeutic Chemical code 14, and cointerventions; timing of follow‐up assessments; pain outcomes assessed (converted to a 0–100 scale); frequency of adverse events (any adverse events and serious adverse events); results (timing of follow‐up, mean and standard deviation for pain outcomes).
Two independent reviewers extracted the data and any disagreements were resolved by discussion or arbitration by a third reviewer when consensus could not be reached. The estimates of treatment effect from trials employing enriched vs. nonenriched designs were considered and compared within each outcome. The trial authors were contacted if any of these data were missing from the trial report or required clarification. If authors did not answer any of the contact attempts and the effect size was not possible to be estimated (i.e. from graphs, P‐values, standard errors), we excluded the trial from the analysis.
The time periods for classifying the follow‐up times were: short‐term (<3 months after randomization), intermediate (≥3 months but <12 months after randomization) and long‐term (12 months or more after randomization). When there were multiple time points that fell within the same category the one that was closest to 6 weeks (short‐term), 6 months (intermediate) and 12 months (long‐term) was used.
The assessment of methodological quality of included trials was performed using the 10‐item Physiotherapy Evidence Database (PEDro) scale (random allocation, concealed allocation, baseline comparability, blinding of subjects, blinding of therapists, blinding of assessors, adequate follow‐up, intention‐to‐treat analysis, between‐group comparisons, and point estimates and variability). Two authors independently assessed methodological quality and any disagreements were resolved by discussion or arbitration by a third reviewer when consensus could not be reached. We used the PEDro scale as it has been found to have acceptable reliability 15 and validity 16, 17.
Data synthesis and analysis
The first stage of the analysis was to establish drug effect sizes by calculating the mean between‐group difference for each trial and then combining these data in meta‐analyses 18. Between‐trial heterogeneity was identified using the I2 statistic, and random effect models were used to account for between‐trial heterogeneity where appropriate. After calculating meta‐analyses with individual trial data, we calculated the difference in effect sizes between enriched and nonenriched designs. For pain, the difference was expressed using mean difference (MD) and 95% confidence interval (CI) on a 0–100 pain scale. For adverse events, we used the relative risk (RR) and 95% CI.
The second stage of the analysis was to conduct meta‐regression analyses to investigate possible sources of heterogeneity in effect size among included trials for the primary outcome. The trial effect size (mean between‐group difference for pain) was the dependent variable, and the independent variables were the design (enriched and nonenriched trials), opioid dosage in morphine equivalent units (for opioids only), and methodological features (allocation concealment, blinding of assessors and intention‐to‐treat analysis). We first conducted the univariate metaregressions to assess the association between each variable and the effect sizes, in which the strongest predictors (P<0.2) were entered into a multivariate model and a backward stepwise approach was used to remove covariates that did not contribute significantly to the model. The best model was identified by examining the proportion of overall between‐trial variability explained by each model (as assessed using the tau statistic) 19.
The analyses were conducted separately for each type of drug (opioids and other analgesics) and each follow‐up time (short‐, intermediate‐ and long‐term). Review Manager (RevMan) software (version 5) was used for the meta‐analyses. Random‐effects meta‐regression was conducted using the metareg command in STATA 14 software and weighted using effect size standard errors. A sensitivity analysis of the type of enriched design (complete or partial) was planned but could not be conducted because it was not possible to categorize the included enriched trials into partial or complete enrichment due to the lack of data provided.
Results
The searches yielded 3766 references, and, after the screening of title, abstracts and full texts we included 108 randomized controlled trials (102 full text articles) that fulfilled the inclusion criteria (Figure 1). Six full text articles included data from more than one trial and each trial was included separately 20, 21, 22, 23, 24, 25. Forty‐two trials were considered enriched designs and 66 were nonenriched designs. Nine trials (eight articles) were not included in the analysis: seven trials (six articles) did not provide sufficient data and the authors did not reply to our contact attempts 25, 26, 27, 28, 29, 30; one trial 24 provided the same data as another trial already included in the analysis; and one trial did not use a continuous scale to assess pain intensity 31. Of the 99 trials included in the analysis (n = 44 171), 32 evaluated opioid analgesics (72 comparisons) and 69 evaluated other types of analgesics (137 comparisons). Some trials evaluated both opioids and other analgesics in different groups and were included in both comparisons. The included trials only reported short‐term and intermediate‐term follow‐ups, so no analyses were performed for long‐term follow‐up. Table 1 presents the details of the included trials. The characteristics of each trial included in this analysis are available in Appendix S2.
Figure 1.
Flowchart of the inclusion process of the study
Table 1.
Characteristics of included trials (n = 108)
| Sample size, median (IQR) | 291 (99–557) |
| Mean age, median (IQR), years | 59.3 (53.5–62.2) |
| Total PEDro score, median (IQR) | 7.0 (7.0–8.0) |
| Design, n (%) | |
| Nonenriched | 66 (61) |
| Enriched | 42 (39) |
| Condition, n (%) | |
| Spinal pain | 24 (23.5) |
| Osteoarthritis (hip or knee) | 78 (76.5) |
IQR, interquartile range.
The difference in the effect estimates between enriched and nonenriched designs for both pain and adverse events are presented in Figure 2 and Figure 3. There were no statistically significant differences in the effect size estimates for pain intensity between enriched and nonenriched designs for opioid analgesics in the short‐term (difference in MD –0.9, 95% CI –2.9 to 1.2), or intermediate‐term (difference in MD 1.3, 95% CI –3.2 to 5.7), and other analgesics in the short‐term (difference in MD 0.8, 95% CI –2.0 to 3.6). We found a significant difference favouring an enriched design (i.e. higher pain reduction) for other analgesics at intermediate‐term, but the effect size was small (difference in MD –4.5, 95% CI –6.4 to –2.5).
Figure 2.
Difference in the effect estimates between enriched and nonenriched designs for pain intensity
Figure 3.
Difference in the effect estimates between enriched and nonenriched designs for any or serious adverse events
For any adverse events, there was a statistically significant difference between enriched and nonenriched designs for opioids (ratio of RR 0.88, 95% CI 0.79 to 0.98), but no significant difference for other analgesics (ratio of RR 0.96, 95% CI 0.90 to 1.03). For serious adverse events, there were no differences due to design for opioids (ratio of RR 1.08, 95% CI 0.19 to 6.13) or other analgesics (ratio of RR 0.67, 95% CI 0.42 to 1.07).
The results of our metaregression analysis are presented in Table 2. For the opioid analgesic trials, there was no association between the effect size and enriched design, but the effect size decreased by six points (out of 100 points) for trials with an intention‐to‐treat analysis (P = 0.001) at short‐term follow‐up. For the analysis of other analgesics in the short‐term, the effect size was associated with blinding of the assessor, with effect size decreasing by six points for trials with blinding of the assessor. For trials with intermediate‐term follow‐up, the effect size was influenced by the design and intention‐to‐treat analysis. Trials with an enriched design had a four point higher effect size, and trials with intention‐to‐treat analysis had a five point lower effect size.
Table 2.
Metaregression analyses on the association between the effect size (pain) and covariates
| Univariate analysis | Multivariate analysis | |||
|---|---|---|---|---|
| Regression coefficient (95% CI) | P value | Regression coefficient (95% CI) | P value | |
| Opioid analgesics short‐term | ||||
| Enriched study design | –0.58 (–3.97 to 2.81) | 0.73 | – | – |
| Dose (mg MEQ log) | –4.11 (–10.63 to 2.42) | 0.21 | – | – |
| Allocation concealment | –1.45 (–5.13 to 2.23) | 0.43 | – | – |
| Blinding of assessor | –6.28 (–14.10 to 1.54) | 0.11* | –4.95 (–11.49 to 1.59) | 0.13 |
| Intention‐to‐treat analysis | 6.22 (3.05 to 9.40) | <0.001* | 5.92 (2.81 to 9.04) | 0.001* |
| Opioid analgesics intermediate‐term | ||||
| Enriched design | 1.19 (–2.93 to 5.31) | 0.56 | – | – |
| Dose (mg MEQ log) | –5.49 (–12.85 to 1.87) | 0.14* | –6.12 (–13.37 to 1.13) | 0.10 |
| Allocation concealment | 1.08 (–3.49 to 5.64) | 0.64 | – | – |
| Blinding of assessor | 3.24 (–1.49 to 7.96) | 0.17* | 3.66 (–0.98 to 8.30) | 0.12 |
| Intention‐to‐treat analysis | –0.92 (–5.12 to 3.28) | 0.66 | – | – |
| Other analgesics short‐term | ||||
| Enriched study design | –0.97 (–3.81 to 5.76) | 0.69 | ||
| Allocation concealment | –2.82 (–10.68 to 5.03) | 0.48 | ||
| Blinding of assessor | 5.65 (0.94 to 10.36) | 0.02* | 6.17 (1.52 to 10.82) | 0.01* |
| Intention‐to‐treat analysis | 3.52 (–1.38 to 8.43) | 0.16* | 4.32 (–0.38 to 9.01) | 0.07 |
| Other analgesics intermediate‐term | ||||
| Enriched study design | –4.35 (–7.92 to –0.78) | 0.02* | –4.01 (–7.38 to –0.63) | 0.02* |
| Allocation concealment | –1.13 (–5.23 to 2.98) | 0.59 | ||
| Blinding of assessor | –2.17 (–5.80 to 1.46) | 0.24 | ||
| Intention‐to‐treat analysis | 5.60 (1.64 to 9.56) | 0.01* | 5.09 (1.31 to 8.88) | 0.01* |
CI, confidence interval; MEQ, morphine equivalent
A negative coefficient means that trials with that feature had greater effect on pain reduction.
P < 0.2
Discussion
This study found there is little or no influence of enriched study designs on estimates of treatment effect size for pain intensity for opioid and nonopioid analgesics in chronic musculoskeletal pain. There was some evidence that enriched designs reduce the risk of adverse events of any nature, but the results for serious adverse events were imprecise and inconsistent, which prevent strong conclusions. Additionally, there was an association between the treatment effect size (pain intensity) and the methodological quality, with failure to blind the outcome assessor and failure to perform an intention‐to‐treat analysis being associated with larger effect sizes.
The strengths of this review include the prespecified methods, which were registered prospectively on PROSPERO. We confined our review to reports of randomised controlled trials that involved a placebo comparison to focus on least‐biased trials. To our knowledge, this is the first systematic review to evaluate the effect of enrichment design on trials of any analgesics for the main chronic musculoskeletal conditions (spinal pain and osteoarthritis). Another strength is that all included trials used the recommended daily dosages, which makes our sample more homogenous and differences in dose could not account for the differences in effect sizes among trials. Additionally, we included a large sample of 99 trials, including 44 171 participants in the meta‐analysis.
This study has some limitations. First, while we searched for trial reports using the three main databases, we restricted our inclusion to full text articles in English, Portuguese, Dutch or Spanish over a 10‐year period, so the inclusion of trials published in other languages or unpublished data may have been missed. Second, a sensitivity analysis on the type of enriched study design (complete or partial) was planned, but this could not be performed because reporting of the type of enriched design was insufficient to allow classification. Finally, we cannot generalize the findings of this study to infer about effectiveness or efficacy of the drugs included since this is a methodological study that aimed comparing two different designs (enriched vs. nonenriched) in pharmacological clinical trials.
Previous comparisons of effects sizes in enriched and nonenriched study designs have produced inconsistent results. Trijau et al. 10 found that trials that employed a flare study design reported larger treatment effects than nonflare designs (effect sizes: 0.66 vs. 0.45, respectively for pain intensity, P<0.0001, 33 studies, 20 915 patients) in trials evaluating NSAIDs for hip, knee and hand osteoarthritis. In contrast, Furlan et al. 3, compared enriched and nonenriched trials of opioid analgesics for chronic noncancer pain and found that enriched study designs do not appear to influence the estimates of treatment efficacy (effect sizes: 0.62 vs. 0.57, respectively for pain intensity, P = 0.6, 62 studies, 11 927 participants), but less adverse events for enriched designs (eight adverse events reported in enriched designs against 26 in nonenriched designs). However, none of these previous studies have adjusted the analysis for methodological quality, a factor that is known to influence the effect estimates and is likely to explain the variation between the studies 8, 32, 33, 34. Another explanation for the differing results is that these previous studies included different types of patients, drug types, and types of enriched designs (e.g. responders and flare designs).
Another finding of this study is the association between the treatment effect size and some methodological features. Failure to blind the outcome assessor and failure to perform an intention‐to‐treat analysis were associated with larger effect sizes. This result is not surprising since poor methodological quality in randomized controlled trials is associated with an overestimation of treatment effects 35. Moreover, a recent study has investigated the source of bias in enriched enrolment studies and reported blinding as one of the major methodological limitations in this design 36. While intention‐to‐treat analysis as a potential source of bias was not investigated, we found that this could influence the effect estimates for either opioids or other analgesics.
While the use of an enriched design was hypothesized to increase the estimated effect of an intervention and reduce adverse events compared with nonenriched designs, we did not find clear differences. Only one in four of our comparisons had significantly larger effect size for an enriched design in pain intensity (i.e. NSAIDs, paracetamol in the intermediate term), but this difference was small and probably not clinically important. We found some evidence that enriched designs reduce the risk of any adverse events, but not serious events. Thus, it is difficult to say that clinicians would find less adverse events if treating only responders to the drug. However, further research is needed to confirm the usefulness of enriched designs for reducing the number of adverse events in chronic pain trials. It seems that it is not the study design (enriched or nonenriched designs) that leads to different treatment effect estimates but other aspects of the trial quality that co‐occur in these trials (e.g. blinding of assessor and intention‐to‐treat analysis). Clinicians should be aware that features of methodological quality may be more important when selecting evidence than trial enrolment design.
In conclusion, there is little or no influence of enriched designs on the estimates of treatment effect for pain intensity. There is some evidence that clinical trials that employ enriched study designs report a reduced risk of adverse events in trials for chronic musculoskeletal pain, but it is unclear whether enriched designs influence estimates of serious adverse events. Importantly, features of trial design such as assessor blinding and intention‐to‐treat analysis were associated with effect estimates for pain intensity in chronic pain trials.
Nomenclature of targets and ligands
Key ligands in this article are hyperlinked to corresponding entries in http://www.guidetopharmacology.org, the common portal for data from the IUPHAR/BPS Guide to PHARMACOLOGY 37.
Competing Interests
A.J.M., C.G.M. and C.‐W.C.L. are investigators on the following three trials evaluating medicines for low back pain.
PACE was an investigator‐initiated trial evaluating paracetamol for acute low back pain. It was funded by the National Health and Medical Research Council of Australia and GlaxoSmithKline. ACTRN 12609000966291
PRECISE is an investigator‐initiated trial evaluating pregabalin for sciatica. It is funded by the National Health and Medical Research Council of Australia with in kind research support from Pfizer. ACTRN12613000530729
OPAL is an investigator‐initiated trial evaluating an opioid analgesic for acute low back pain funded by the National Health and Medical Research Council of Australia. ACTRN12615000775516
T.P.Y. is supported by CAPES (Coordination for the Improvement of Higher Education Personnel), Ministry of Education, Brazil. B.T.S. is supported by CNPq (National Council for Scientific and Technological Development), Ministry of Science and Technology, Brazil. C.G.M. is supported by a Principal Research Fellowship from the National Health and Medical Research Council, Australia. A.J.M. is the Program Director of the NHMRC Centre for Research Excellence on Medicines and Ageing. C.‐W.C.L. is supported by a Career Development Fellowship from the National Health and Medical Research Council, Australia.
Supporting information
Supporting info item
Figure S1 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating opioids at intermediate term
Figure S2 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating opioids at short term
Figure S3 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating other analgesics at intermediate term
Figure S4 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating other analgesics at short term
Appendix S1 Search strategy for Medline database
Appendix S2 Characteristics of included studies (n = 108)
Text S1 Checklist of items to include when reporting a systematic review or meta‐analysis
Yamato, T. P. , Maher, C. G. , Saragiotto, B. T. , Shaheed, C. A. , Moseley, A. M. , Lin, C.‐W. C. , Koes, B. , and McLachlan, A. J. (2017) Comparison of effect sizes between enriched and nonenriched trials of analgesics for chronic musculoskeletal pain: a systematic review. Br J Clin Pharmacol, 83: 2347–2355. doi: 10.1111/bcp.13350.
References
- 1. Ioannidis JP, Greenland S, Hlatky MA, Khoury MJ, Macleod MR, Moher D, et al. Increasing value and reducing waste in research design, conduct, and analysis. Lancet 2014; 383: 166–175. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 2. Straube S, Derry S, McQuay HJ, Moore RA. Enriched enrollment: definition and effects of enrichment and dose in trials of pregabalin and gabapentin in neuropathic pain. A systematic review. Br J Clin Pharmacol 2008; 66: 266–275. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Furlan A, Chaparro LE, Irvin E, Mailis‐Gagnon A. A comparison between enriched and nonenriched enrollment randomized withdrawal trials of opioids for chronic noncancer pain. Pain Res Manag 2011; 16: 337–351. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Katz N. Enriched enrollment randomized withdrawal trial designs of analgesics: focus on methodology. Clin J Pain 2009; 25: 797–807. [DOI] [PubMed] [Google Scholar]
- 5. Lemmens HJ, Wada DR, Munera C, Eltahtawy A, Stanski DR. Enriched analgesic efficacy studies: an assessment by clinical trial simulation. Contemp Clin Trials 2006; 27: 165–173. [DOI] [PubMed] [Google Scholar]
- 6. Temple R. Enrichment of clinical study populations. Clin Pharmacol Ther 2010; 88: 774–778. [DOI] [PubMed] [Google Scholar]
- 7. Abdel Shaheed C, Maher CG, Williams KA, Day R, McLachlan AJ. Efficacy, tolerability, and dose‐dependent effects of opioid analgesics for low back pain: a systematic review and meta‐analysis. JAMA Intern Med 2016; 176: 958–968. [DOI] [PubMed] [Google Scholar]
- 8. Schulz KF, Chalmers I, Hayes RJ, Altman DG. Empirical evidence of bias. Dimensions of methodological quality associated with estimates of treatment effects in controlled trials. JAMA 1995; 273: 408–412. [DOI] [PubMed] [Google Scholar]
- 9. McQuay HJ, Derry S, Moore RA, Poulain P, Legout V. Enriched enrolment with randomised withdrawal (EERW): time for a new look at clinical trial design in chronic pain. Pain 2008; 135: 217–220. [DOI] [PubMed] [Google Scholar]
- 10. Trijau S, Avouac J, Escalas C, Gossec L, Dougados M. Influence of flare design on symptomatic efficacy of non‐steroidal anti‐inflammatory drugs in osteoarthritis: a meta‐analysis of randomized placebo‐controlled trials. Osteoarthritis Cartilage 2010; 18: 1012–1018. [DOI] [PubMed] [Google Scholar]
- 11. Airaksinen O, Brox JI, Cedraschi C, Hildebrandt J, Klaber‐Moffett J, Kovacs F, et al. Working group on guidelines for chronic low back pain. Chapter 4. European guidelines for the management of chronic nonspecific low back pain. Eur Spine J 2006; 15 (Suppl 2): S192–S300. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12. Korff MV, Saunders K, Thomas Ray G, Boudreau D, Campbell C, Merrill J, et al. De facto long‐term opioid therapy for noncancer pain. Clin J Pain 2008; 24: 521–527. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13. Holford NH, Sheiner LB. Understanding the dose‐effect relationship: clinical application of pharmacokinetic‐pharmacodynamic models. Clin Pharmacokinet 1981; 6: 429–453. [DOI] [PubMed] [Google Scholar]
- 14. International language for drug utilisation research. WHO Collaborating Centre for Drug Statistics Methodology 2011. Available at http://www.whocc.no/atc_ddd_index/ (last accessed July 2016).
- 15. Maher CG, Sherrington C, Herbert RD, Moseley AM, Elkins M. Reliability of the PEDro scale for rating quality of randomized controlled trials. Phys Ther 2003; 83: 713–721. [PubMed] [Google Scholar]
- 16. Macedo LG, Elkins MR, Maher CG, Moseley AM, Herbert RD, Sherrington C. There was evidence of convergent and construct validity of physiotherapy evidence database quality scale for physiotherapy trials. J Clin Epidemiol 2010; 63: 920–925. [DOI] [PubMed] [Google Scholar]
- 17. de Morton NA. The PEDro scale is a valid measure of the methodological quality of clinical trials: a demographic study. Aust J Physiother 2009; 55: 129–133. [DOI] [PubMed] [Google Scholar]
- 18. Herbert RD. How to estimate treatment effects from reports of clinical trials. I: continuous outcomes. Aust J Physiother 2000; 46: 229–235. [DOI] [PubMed] [Google Scholar]
- 19. Pengel LH, Herbert RD, Maher CG, Refshauge KM. Acute low back pain: systematic review of its prognosis. BMJ 2003; 327: 323. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20. Bingham CO 3rd, Buckland‐Wright JC, Garnero P, Cohen SB, Dougados M, Adami S, et al. Risedronate decreases biochemical markers of cartilage degradation but does not decrease symptoms or slow radiographic progression in patients with medial compartment osteoarthritis of the knee: results of the two‐year multinational knee osteoarthritis structural arthritis study. Arthritis Rheum 2006; 54: 3494–3507. [DOI] [PubMed] [Google Scholar]
- 21. Bingham CO 3rd, Sebba AI, Rubin BR, Ruoff GE, Kremer J, Bird S, et al. Efficacy and safety of etoricoxib 30 mg and celecoxib 200 mg in the treatment of osteoarthritis in two identically designed, randomized, placebo‐controlled, non‐inferiority studies. Rheumatology 2007; 46: 496–507. [DOI] [PubMed] [Google Scholar]
- 22. Birbara C, Ruoff G, Sheldon E, Valenzuela C, Rodgers A, Petruschke RA, et al. Efficacy and safety of rofecoxib 12.5 mg and celecoxib 200 mg in two similarly designed osteoarthritis studies. Curr Med Res Opin 2006; 22: 199–210. [DOI] [PubMed] [Google Scholar]
- 23. Hochberg MC, Fort JG, Svensson O, Hwang C, Sostek M. Fixed‐dose combination of enteric‐coated naproxen and immediate‐release esomeprazole has comparable efficacy to celecoxib for knee osteoarthritis: two randomized trials. Curr Med Res Opin 2011; 27: 1243–1253. [DOI] [PubMed] [Google Scholar]
- 24. Skljarevski V, Zhang S, Chappell AS, Walker DJ, Murray I, Backonja M. Maintenance of effect of duloxetine in patients with chronic low back pain: a 41‐week uncontrolled, dose‐blinded study. Pain Med 2010; 11: 648–657. [DOI] [PubMed] [Google Scholar]
- 25. Schiff M, Minic M. Comparison of the analgesic efficacy and safety of nonprescription doses of naproxen sodium and ibuprofen in the treatment of osteoarthritis of the knee. J Rheumatol 2004; 31: 1373–1383. [PubMed] [Google Scholar]
- 26. Bingham CO 3rd, Smugar SS, Wang H, Peloso PM, Gammaitoni A. Predictors of response to cyclo‐oxygenase‐2 inhibitors in osteoarthritis: pooled results from two identical trials comparing etoricoxib, celecoxib, and placebo. Pain Med 2011; 12: 352–361. [DOI] [PubMed] [Google Scholar]
- 27. Chopra A, Lavin P, Patwardhan B, Chitre D. A 32‐week randomized, placebo‐controlled clinical evaluation of RA‐11, an Ayurvedic drug, on osteoarthritis of the knees. J Clin Rheumatol 2004; 10: 236–245. [DOI] [PubMed] [Google Scholar]
- 28. Jamison RN, Edwards RR, Liu X, Ross EL, Michna E, Warnick M, et al. Relationship of negative affect and outcome of an opioid therapy trial among low back pain patients. Pain Pract 2013; 13: 173–181. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29. Luyten FP, Geusens P, Malaise M, De Clerck L, Westhovens R, Raeman F, et al. A prospective randomised multicentre study comparing continuous and intermittent treatment with celecoxib in patients with osteoarthritis of the knee or hip. Ann Rheum Dis 2007; 66: 99–106. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 30. Sawitzke AD, Shi H, Finco MF, Dunlop DD, Harris CL, Singer NG, et al. Clinical efficacy and safety of glucosamine, chondroitin sulphate, their combination, celecoxib or placebo taken to treat osteoarthritis of the knee: 2‐year results from GAIT. Ann Rheum Dis 2010; 69: 1459–1464. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31. Golden HE, Moskowitz RW, Minic M. Analgesic efficacy and safety of nonprescription doses of naproxen sodium compared with acetaminophen in the treatment of osteoarthritis of the knee. Am J Ther 2004; 11: 85–94. [DOI] [PubMed] [Google Scholar]
- 32. Gluud LL. Bias in clinical intervention research. Am J Epidemiol 2006; 163: 493–501. [DOI] [PubMed] [Google Scholar]
- 33. Kjaergard LL, Villumsen J, Gluud C. Reported methodologic quality and discrepancies between large and small randomized trials in meta‐analyses. Ann Intern Med 2001; 135: 982–989. [DOI] [PubMed] [Google Scholar]
- 34. Moher D, Pham B, Jones A, Cook DJ, Jadad AR, Moher M, et al. Does quality of reports of randomised trials affect estimates of intervention efficacy reported in meta‐analyses? Lancet 1998; 352: 609–613. [DOI] [PubMed] [Google Scholar]
- 35. Savovic J, Jones H, Altman D, Harris R, Juni P, Pildal J, et al. Influence of reported study design characteristics on intervention effect estimates from randomised controlled trials: combined analysis of meta‐epidemiological studies. Health Technol Assess 2012; 16: 1–82. [DOI] [PubMed] [Google Scholar]
- 36. Moore RA, Wiffen PJ, Eccleston C, Derry S, Baron R, Bell RF, et al. Systematic review of enriched enrolment, randomised withdrawal trial designs in chronic pain: a new framework for design and reporting. Pain 2015; 156: 1382–1395. [DOI] [PubMed] [Google Scholar]
- 37. Southan C, Sharman JL, Benson HE, Faccenda E, Pawson AJ, Alexander SPH, et al. The IUPHAR/BPS Guide to PHARMACOLOGY in 2016: towards curated quantitative interactions between 1300 protein targets and 6000 ligands. Nucleic Acids Res 2016; 44: D1054–D1068. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supporting info item
Figure S1 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating opioids at intermediate term
Figure S2 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating opioids at short term
Figure S3 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating other analgesics at intermediate term
Figure S4 Forest plot of the effect estimates for enriched and non‐enriched trials evaluating other analgesics at short term
Appendix S1 Search strategy for Medline database
Appendix S2 Characteristics of included studies (n = 108)
Text S1 Checklist of items to include when reporting a systematic review or meta‐analysis