quinta-feira, 11 de junho de 2015

Compliance effects in a randomised controlled trial of yoga for chronic low back pain: a methodological study

Open Access funded by Arthritis Research UK

Abstract

Study design

Methodological study nested within a multicentre randomised controlled trial (RCT) of yoga plus usual general practitioner (GP) care vs usual GP care for chronic low back pain.

Objective

To explore the treatment effects of non-compliance using three approaches in an RCT evaluating yoga for low back pain.

Summary of background data

A large multicentre RCT using intention-to-treat (ITT) analysis found that participants with chronic low back pain who were offered a 12-week progressive programme of yoga plus usual GP care had better back function than those offered usual GP care alone. However, ITT analysis can underestimate the effect of treatment in those who comply with treatment. As such, the data were analysed using other approaches to assess the problem of non-compliance. The main outcome measure was the self-reported Roland Morris Disability Questionnaire (RMDQ).

Methods

Complier average causal effect (CACE) analysis, per-protocol analysis and on-treatment analysis were conducted on the data of participants who were fully compliant, predefined as attendance of at least three of the first six sessions and at least three other sessions. The analysis was repeated for participants who had attended at least one yoga session (i.e. any compliance), which included participants who were fully compliant. Each approach was described, including strengths and weaknesses, and the results of the different approaches were compared with those of the ITT analysis.

Results

For the participants who were fully compliant (n = 93, 60%), a larger beneficial treatment effect was seen using CACE analysis compared with per-protocol, on-treatment and ITT analyses at 3 and 12 months. The difference in mean change in RMDQ score between randomised groups was −3.30 [95% confidence interval (CI) −4.90 to −1.70, P < 0.001] at 3 months and −2.23 (95% CI −3.93 to −0.53, P = 0.010) at 12 months for CACE analysis, −3.12 (95% CI −4.26 to −1.98, P < 0.001) at 3 months and −2.11 (95% CI −3.33 to −0.89, P = 0.001) at 12 months for per-protocol analysis, and −2.91 (95% CI −4.06 to −1.76, P < 0.001) at 3 months and −2.10 (95% CI −3.31 to −0.89, P = 0.001) at 12 months for on-treatment analysis. For the participants who demonstrated any compliance (n = 133, 85%), the results were generally consistent with the fully compliant group at 3 months, but the treatment effect was smaller. The difference in mean change in RMDQ score between randomised groups was −2.45 (95% CI −3.67 to −1.24) for CACE analysis, −2.30 (95% CI −3.43 to 1.17) for per-protocol analysis and −2.15 (95% CI −3.25 to −1.06) for on-treatment analysis, which was slightly less than that for ITT analysis. In contrast, at 12 months, per-protocol and on-treatment analyses showed a larger treatment effect compared with CACE and ITT analyses: per protocol analysis −1.86 (95% CI −3.02 to −0.71), on-treatment analysis −1.99 (95% CI −3.13 to −0.86) and CACE analysis −1.67 (95% CI −2.95 to −0.40).

Conclusion

ITT analysis estimated a slightly smaller treatment effect in participants who complied with treatment. When examining compliance, CACE analysis is more rigorous than per-protocol and on-treatment analyses. Using CACE analysis, the treatment effect was larger in participants who complied with treatment compared with participants who were allocated to treatment, and the difference between ITT and CACE analyses for the fully compliant group at 3 months was small but clinically important. Per-protocol and on-treatment analyses may produce unreliable estimates when the effect of treatment is small.

International Standard Randomised Trial Number Register

ISRCTN 81079604.

Keywords

  • Compliance effects; 
    •  
  • CACE analysis; 
    •  
  • Per-protocol analysis; 
    •  
  • On-treatment analysis; 
    •  
  • Yoga; 
    •  
  • Low back pain

Introduction

Intention-to-treat (ITT) analysis is widely recommended and the accepted method for analysing outcome data in randomised controlled trials. In ITT analysis, all patient data are analysed in the original group to which they were randomised, irrespective of whether or not they completed the full treatment or intervention on offer. ITT analysis therefore allows comparison of the effectiveness of treatment(s) offered rather than treatment(s) received, and is useful to policy makers who need to make a decision about whether to make a treatment available based on its effectiveness. ITT analysis reflects what would actually happen in clinical practice. However, if patients do not complete the full treatment, ITT analysis underestimates the treatment effect in patients who complied, and therefore does not answer the question ‘What is the effect of the treatment in individuals who comply with treatment?’. Other analyses may be considered to address the problem of non-compliance in trials [1].
Per-protocol analysis only compares the data for participants who complied with their randomised treatment (i.e. data for patients who did not comply are excluded). On-treatment analysis compares the data for participants by the treatment they received, regardless of the group to which they were randomised. In both these analyses, the effects of randomisation are lost, which can potentially introduce bias and consequently threaten the internal validity of the results. Consolidated Standards of Reporting Trials (CONSORT) [1] advises that, in cases where per-protocol analysis is reported, it should be described as a ‘non-randomised, observational comparison’ [2] and [3]. On-treatment analysis is not discussed in CONSORT, but the present authors believe that the principles for per-protocol analysis should apply.
An alternative to per-protocol and on-treatment analyses, which respects randomisation, is complier average causal effect (CACE) analysis. CACE analysis is a measure of the causal effect of a treatment or intervention on the people who received it as intended by the original group allocation. In this analysis, compliers in the intervention group are compared with a like-for-like group in the control group. As CACE analysis retains the initial randomisation, it overcomes the problems faced by per-protocol and on-treatment analyses.
A recent randomised trial of yoga for chronic low back pain, conducted by the authors, compared 12 classes of a specialised programme of yoga plus usual general practitioner (GP) care with usual GP care[4]. One hundred and fifty-six participants were randomised to yoga plus usual GP care and 157 participants were randomised to usual GP care alone. Full compliance was predefined as attendance of at least three of the first six classes and at least three other classes [5]. Ninety-three (60%) participants were fully compliant, 40 (26%) participants attended at least one class but did not meet the definition of full compliance, and 23 (15%) participants did not attend any classes. Outcomes were analysed and reported using ITT analysis [4], but given that 40% of the participants did not comply with the intervention, it was considered that this approach provided a diluted estimate. The aim of the present study was to explore the effect of non-compliance in the yoga trial, and compare the treatment effect using three different analyses: per-protocol, on-treatment and CACE analyses. Observational comparisons were included as recent trials have reported using per-protocol analysis [6] and [7]. The results of these analyses were compared with the results of the ITT analysis from the main trial to determine how much ITT analysis underestimated the treatment effect. The yoga trial data provided the opportunity to explore how the data responded under different analyses. It was predicted that CACE analysis would provide a more reliable and conservative estimate of treatment effect than per-protocol and on-treatment analyses, as random allocation is ignored in the latter two types of analysis.

Methods

Full details of the trial have been reported elsewhere [4]. Briefly, participants were recruited from primary care from five areas of the UK. Eligible participants were randomised equally to receive 12 classes of a specialised programme of yoga plus usual GP care or usual GP care alone. All participants remained under the care of their GP and received usual National Health Service treatments, such as GP consultations, prescribed painkillers and referral to physiotherapy [8]. Both groups received a copy of ‘The back book’ [9]. Eligible participants completed postal questionnaires at baseline, 3 months (immediately post intervention), 6 months and 12 months. The main outcome was back function, measured using the Roland Morris Disability Questionnaire (RMDQ) [10] at 3-month follow-up. Data on class attendance were collected from class registers, and details of home practice were collected from the postal questionnaires. Allowing for a 20% loss to follow-up, the authors sought to recruit 262 participants (131 in each group) [11]. In total, 313 participants were recruited.

Definition of full compliance

Full compliance was predefined as attendance of at least three of the first six classes and at least three other classes [11]. At the initial trial design phase, the designer of the ‘Yoga for Healthy Lower Backs’ programme (AT) and the other yoga consultants (AS and JDA) estimated that if participants adhered to these minimum criteria, they could be said to have ‘received sufficient yoga’ as a self-management toolkit that would be likely to show long-term improvements, rather than just having been ‘offered some yoga’ for short-term relief. The course involved 12 classes that gradually and progressively introduced participants to more complex – but nevertheless still gentle and suitable for beginners – yoga by building on previous knowledge and skills. However, the definition was subjective and the exact level at which the programme might be effective could not be known. All participants allocated to the yoga group received a relaxation CD regardless of whether or not they attended a class. Participants who attended at least one class were given a yoga manual with details on how to practice the yoga and how to take care of their backs, and therefore had the opportunity to practice the yoga at home. Therefore, the analysis was re-run using ‘any compliance’, defined as attendance of one or more classes. It was estimated that this would provide a more conservative estimate of treatment effect compared with the participants who were fully compliant. The opportunity for contamination in this trial was minimised as the yoga programme was not available outside of the trial.

Analysis

ITT analysis

The findings from the main trial using an ITT analysis, which are reported here for comparison with the results of the other analyses, were taken directly from the published paper [4]. This analysis was conducted using a linear mixed model comparing changes in RMDQ scores from baseline over time. Time was treated as a categorical variable and was included as a fixed effect in addition to group, age, sex, eligibility RMDQ score, class preference (stratification factor), group and time interaction, and duration of back pain. The ITT analysis compared Group A (all participants allocated to yoga) with Group B (all participants allocated to usual GP care) (see Fig. 1).
Full-size image (12 K)
Fig. 1.
Treatment profile of the yoga trial – fully compliant. GP, general practitioner.
Figure options

Per-protocol and on-treatment analyses

Fully compliant group

For the per-protocol analysis, the 63 participants who did not meet the definition of full compliance, including those who did not attend any yoga sessions, were excluded from the analysis, and the remaining intervention group was compared with the usual GP care group (A1 vs B in Fig. 1). For the on-treatment analysis, the 23 participants who did not attend any yoga sessions and the 40 participants who attended at least one session but were not fully compliant plus the usual GP care group were compared with the remaining intervention group participants (A1 vs A2 + B in Fig. 1).

Any compliance group

For the per-protocol analysis, the 23 participants who did not attend any yoga sessions were excluded from the analysis, and the remaining intervention group was compared with the usual GP care group (A1 vs B inFig. 2). In the on-treatment analysis, the 23 participants who did not attend any yoga sessions plus the usual GP care group were compared with the remaining intervention group participants (A1 vs A2 + B inFig. 2).
Full-size image (9 K)
Fig. 2.
Treatment profile of the yoga trial – any compliance. GP, general practitioner.
Figure options

CACE analysis

For the CACE analysis, an instrumental variable approach was used, with the ivreg command in Stata, StataCorp LP, Texas, USA [12] and [13]. The model was extended to include age, sex, baseline RMDQ score, class preference (stratification factor) and duration of back pain. The analyses were repeated for the change from baseline at 3 and 12 months. For this analysis, it was assumed that members of the control group had the same probability of non-compliance as members of the intervention group, and being offered the treatment had no effect on outcome [14].
The effect of the number of yoga sessions attended was explored in its continuous form with estimation using instrumental variables [15]. Using the continuous compliance data, it was assumed that the causal effect of d sessions is proportional to the number of sessions (e.g. 12 sessions are twice as good as six sessions).

Results

Compliance overview

Of the 156 participants randomised to yoga, 93 (60%) participants attended at least three of the first six sessions and at least three other sessions (i.e. were fully compliant), 40 (26%) participants attended at least one class but were not fully compliant, and 23 (15%) participants did not attend any classes (Table A, see online supplementary material). Of the 156 individuals randomised to yoga, 133 (85%) attended at least one class [93 (60%) + 40 (26%)]. Of the 102 participants who reported practising yoga at home at 3 months and the 74 participants who reported practising yoga at home at 12 months, more than half had attended nine or more classes (Table A, see online supplementary material). Data were missing for the question about home practice, and of the 30 participants who did not answer the question about home practice at 3 months and 30 participants who did not answer the question about home practice at 12 months, over 70% had attended two or fewer classes.

Impact of compliance on treatment estimates

Fully compliant group

The difference in mean change in RMDQ score between the groups with adjustments for non-compliance for CACE analysis was −3.30 (95% CI −4.90 to −1.70, P < 0.001) at 3 months and −2.23 (95% CI −3.93 to −0.53, P = 0.01) at 12 months ( Table 1). The CACE estimates were larger than the ITT estimates, demonstrating a greater benefit of yoga amongst participants who were fully compliant.
Table 1.
Full compliance: results of compliance using different approaches.
ApproachBetween-group difference in mean change in RMDQ score from baseline (95% CI)a
Month 3Month 12
ITT−2.17 (−3.31 to −1.03), P < 0.001−1.57 (−2.71 to −0.42), P = 0.007
CACE−3.30 (−4.90 to −1.70), P < 0.001−2.23 (−3.93 to −0.53), P = 0.010
Per-protocol−3.12 (−4.26 to −1.98), P < 0.001−2.11 (−3.33 to −0.89), P = 0.001
On-treatment−2.91 (−4.06 to −1.76), P < 0.001−2.10 (−3.31 to −0.89), P = 0.001
ITT, intention to treat (Estimate from Tilbrook et al., 2011); CACE, complier average causal effect; CI, confidence interval.
Results are for the Roland Morris Disability Questionnaire (RMDQ).
A minus sign indicates better health in the yoga group. A higher score indicates better health.
a
Adjustments were made for age, sex, baseline score, class preference and duration of back pain.
Table options

Any compliance group

The difference in mean change in RMDQ score between the groups with adjustments for non-compliance for CACE analysis was −2.45 (95% CI −3.67 to −1.24, P < 0·001) at 3 months and −1.67 (95% CI −2.95 to −0.40, P = 0.01) at 12 months ( Table 2). The CACE estimates were larger than the ITT estimates, demonstrating a greater benefit of yoga amongst participants who attended one or more sessions.
Table 2.
Any compliance: results of compliance using different approaches.
ApproachBetween-group difference in mean change in RMDQ score from baseline (95% CI)a
Month 3Month 12
ITT−2.17 (−3.31 to −1.03), P < 0.001−1.57 (−2.71 to −0.42), P = 0.007
CACE−2.45 (−3.67 to −1.24), P < 0.001−1.67 (−2.95 to −0.40), P = 0.01
Per-protocol−2.30 (−3.43 to −1.17), P < 0.001−1.86 (−3.02 to −0.71), P = 0.002
On-treatment−2.15 (−3.25 to −1.06), P < 0.001−1.99 (−3.13 to −0.86), P = 0.001
ITT, intention to treat (Estimate from Tilbrook et al., 2011); CACE, complier average causal effect; CI, confidence interval.
Results are for the Roland Morris Disability Questionnaire (RMDQ).
A minus sign indicates better health in the yoga group. A higher score indicates better health.
a
Adjustments were made for age, sex, baseline score, class preference and duration of back pain.
Table options
Modelling the continuous compliance measure of the number of sessions attended showed that, for each extra session, the mean change in RMDQ score was −0.31 (95% CI −0.46 to −0.16, P < 0.001) at 3 months and −0.21 (95% CI −0.37 to −0.05, P = 0.01) at 12 months.

Comparison of treatment effects using three different analyses for non-compliance

Fully compliant group

The estimated between-group differences in mean change in RMDQ score for ITT, CACE, per-protocol and on-treatment analyses are reported in Table 1 and Fig. 3. At 3 months, the ITT analysis suggested a −2.17 (95% CI −3.31 to −1.03) mean change in RMDQ score from baseline, and the per-protocol, on-treatment and CACE analyses suggested a larger effect: −3.12 (95% CI −4.26 to −1.98), −2.91 (95% CI −4.06 to −1.76) and −3.30 (95% CI −4.90 to −1.70), respectively. The same pattern in results was seen at 12 months; the ITT analysis suggested a −1.57 (95% CI −2.71 to −0.42) mean change in RMDQ score from baseline, and the per-protocol, on-treatment and CACE analyses suggested larger effects: −2.11 (95% CI −3.33 to −0.89), −2.10 (95% CI −3.31 to −0.89) and −2.23 (95% CI −3.93 to −0.53), respectively.
Full-size image (39 K)
Fig. 3.
Forest plot comparing results of intention-to-treat (ITT) analysis with complier average causal effect (CACE), per-protocol and on-treatment analyses.
Figure options

Any compliance group

The estimated between-group differences in mean change in RMDQ score for ITT, CACE, per-protocol and on-treatment analyses are reported in Table 2 and Fig. 3. At 3 months, the ITT analysis suggested a −2.17 (95% CI −3.31 to −1.03) mean change in RMDQ score from baseline, and the per-protocol and CACE analyses suggested a larger effect: −2.30 (95% CI −3.43 to −1.17) and −2.45 (95% CI −3.67 to −1.24), respectively. The on-treatment analysis suggested a similar effect to the ITT analysis [−2.15 (95% CI −3.25 to −1.06)]. A slightly different pattern in results was seen at 12 months; the ITT analysis suggested a −1.57 (95% CI −2.71 to −0.42) mean change in RMDQ score from baseline, and the per-protocol and on-treatment analyses suggested larger effects: −1.86 (95% CI −3.02 to −0.71) and −1.99 (95% CI −3.13 to −0.86), respectively. The CACE estimate was between the ITT, per-protocol and on-treatment estimates at −1.67 (95% CI −2.95 to −0.40).
For CACE analysis, the difference in mean scores between the groups with adjustments for non-compliance was −3.30 (95% CI −4.90 to −1.70, P < 0.001) at 3 months and −2.23 (95% CI −3.93 to −0.53,P = 0.01) at 12 months. When partial attenders were included as compliers, compliance was 133 (85%) and the estimated causal effect was smaller as it was averaged over a larger group (3 months: −2.45, 95% CI −3.67 to −1.24, P < 0.001; 12 months: −1.67, 95% CI −2.95 to −0.40, P = 0.01).

Discussion

ITT analysis consistently estimated a slightly smaller treatment effect of yoga among participants who attended one or more sessions and who were fully compliant compared with CACE, per-protocol and on-treatment analyses, except for one instance; ITT analysis estimated a slightly larger treatment effect compared with on-treatment analysis at 3 months for the participants who demonstrated any compliance, which is contrary to the result expected. The treatment effects for participants who were fully compliant are larger than those for the participants who demonstrated any compliance at both time points and for each analysis, which indicates that those who were fully compliant with the treatment had better outcomes. Indeed, there was a pattern to suggest that, on average, there was an additional beneficial effect of yoga for each class attended. The randomisation process ensures balance across groups, and it was expected that there would be participants in both groups whose backs improved to a smaller or larger degree. In the on-treatment analysis, the data for the individuals who did not comply (i.e. did not attend any classes) were moved to the control group and therefore the balance achieved by the randomisation process was lost. ITT analysis suggests that the treatment effect would be at its greatest immediately after the intervention (3 months). There was a difference in the mean change score of −1.13 between CACE and ITT analyses, favouring CACE analysis, for the participants who were fully compliant at 3 months, which is at the bottom of the range of what is recognised as clinically important [16] and [17]. No other comparisons with ITT analysis with 3- and 12-month data gave clinically important differences.
It was anticipated that the effect sizes calculated using per-protocol and on-treatment analyses would be larger than those using CACE and ITT analyses, as the former two analyses are likely to produce more biased estimates. However, this effect was only observed when the treatment effect was smaller (i.e. in the participants who demonstrated any compliance at 12 months). Indeed, on-treatment analysis was not as predicted for the participants who demonstrated any compliance at 3 months. Whilst the effect size was very close to the other estimates, it was slightly less than that for ITT analysis. Again, this might be explained by a smaller effect size producing a biased estimate.
Apart from the participants who demonstrated any compliance at 12 months, CACE analysis gave a larger treatment effect but the differences between CACE, per-protocol and on-treatment analyses were small. The largest difference in treatment effect was observed between ITT and CACE analyses in participants who were fully compliant, which indicates that ITT analysis estimated a smaller treatment effect in participants who comply and this was only clinically important at 3 months.

Comparison with other trials estimating compliance

A review of the literature revealed few trials evaluating musculoskeletal conditions that reported per-protocol analysis and none that reported CACE analysis. A trial of Iyengar yoga therapy for low back pain undertook a per-protocol analysis in addition to ITT analysis [6]. Per-protocol analysis showed greater improvements in back function, as measured by the Oswald Disability Index, compared with ITT analysis at both follow-ups. A trial of the Alexander technique reported no ‘meaningful change in the results’ when only those participants who adhered were included, but this trial did not describe the analysis nor report the results [16]. A trial comparing manual therapy and exercise therapy in osteoarthritis of the hip or knee conducted per-protocol and ITT analyses. However, uncertainty about the accuracy of the data from the self-reported measure of compliance meant that the authors expressed caution in interpreting the results[7].

Strengths

This was a large trial, and by collecting data on compliance, it was possible to estimate the effect of the intervention for participants who were compliant using different approaches. The study found that the different approaches give different but similar results, and that ITT analysis generally provides a more conservative treatment effect.

Limitations

The study definition of ‘any compliance’ was subjective and very conservative. Participants were considered to be compliant if they had attended at least one session, as all participants attending at least one class would have had the opportunity to practice yoga at home. However, whilst some participants who had attended no sessions did practice yoga at home, the general finding was that the greater number of classes attended, with the exception of those attending 12 classes, the more likely participants were to report practising yoga at home. Therefore, whilst ITT analysis was found to estimate a slightly smaller treatment effect for yoga in participants who attended one or more sessions, the estimates produced by the other analyses varied according to the definition of compliance.
In future, in studies where non-compliance might be an issue, consideration should be given to include CACE analysis. In the present trial, measures were taken to reduce non-compliance, such as asking participants to indicate which classes and times they could attend. However, non-compliance was still observed for a variety of reasons. For future trials of group yoga sessions, trialists may also want to consider inflating power calculations to take account of low compliance, so that a more accurate estimate of treatment effect can be obtained by ITT analysis.

Conclusions and recommendations

When examining the effect of treatment on participants who were compliant, CACE analysis is the only approach that respects randomisation. Per-protocol and on-treatment analyses are more likely to produce biased estimates than CACE analysis, and CACE analysis showed a larger beneficial treatment effect in all but one of the analyses. However, differences in treatment effect between CACE, per-protocol and on-treatment analyses were small. In CACE analysis, the treatment effect was larger in participants who complied with treatment compared with participants allocated to treatment; and the difference between ITT and CACE analyses at 3 months for participants who were fully compliant was clinically important, although small. Per-protocol and on-treatment analyses may produce unreliable estimates when the effect of treatment is small. CACE analysis could be considered in trials in which compliance with treatment is low. In anticipation of such cases, trialists should consider a definition of compliance at the start of the trial.

Acknowledgement

The authors wish to thank Helen Cox for her contribution to the original trial protocol and contribution of her work as one of the trial managers.
Ethical approval: Leeds (East) Research Ethics Committee (Ref. No. 07/Q1206/35).
Funding: Arthritis Research UK.
Conflict of interest: Two of the authors (A.S. and A.T.) are self-employed yoga teachers who run a social enterprise and receive royalties from the yoga compact disc and yoga manual, and teach the programme to other yoga teachers. A.T. designed the yoga programme and taught the intervention to the other yoga teachers participating in the trial. AT and AS held some of the yoga classes delivered to participants. They were not involved in patient selection, data acquisition, data analysis or data interpretation in this study or the original trial.
Table S1.  
Number of yoga classes attended and home practice.
Help with DOCX files
Options

References

    • [3]
    • D. Moher, S. Hopewell, K.F. Schulz, V. Montori, P.C. Gøtzsche, P.J. Devereaux, et al.
    • CONSORT 2010 explanation and elaboration: updated guidelines for reporting parallel group randomised trials
    • Br Med J, 15 (2010), p. 340 c869
    • [7]
    • J.H. Abbott, M.C. Robertson, C. Chapple, D. Pinto, A.A. Wright, S. Leon de la Barra, et al.
    • Manual therapy, exercise therapy, or both, in addition to usual care, for osteoarthritis of the hip or knee: a randomized controlled trial. 1: clinical effectiveness
    • Osteoarthritis Cartilage, 21 (2013), pp. 525–534
    • Article
       | 
       PDF (614 K)
       | 
      View Record in Scopus
       | 
      Citing articles (21)
    • [9]
    • K. Burton, J. Klaber Moffett, C. Main, M.G.W. Roland
    • The back book
    • (2nd ed.)The Stationery Office, Norwich (2002)
    • [14]
    • C.E. Hewitt, D.J. Torgerson, J.N.V. Miles
    • Is there another way to take account of noncompliance in randomized controlled trials?
    • Can Med Assoc J, 175 (2006), p. 348
    • [16]
    • P. Little, G. Lewith, F. Webley, M. Evans, A. Beattie, K. Middleton, et al.
    • Randomised controlled trial of Alexander technique lessons, exercise, and massage (ATEAM) for chronic and recurrent back pain
    • Br Med J (2008), p. 337 a884
    • [17]
    • S.E. Lamb, Z. Hansen, R. Lall, E. Castelnuovo, E.J. Withers, V. Nichols, et al.
    • Group cognitive behavioural treatment for low-back pain in primary care: a randomised controlled trial and cost-effectiveness analysis
    • Lancet, 375 (2010), pp. 916–923
    • Article
       | 
       PDF (253 K)
       | 
      View Record in Scopus
       | 
      Citing articles (121)
Corresponding author at: York Trials Unit, Lower Ground Floor, ARRC Building, Department of Health Sciences, University of York, Heslington, York YO10 5DD, UK. Tel.: +44 01904 321668; fax: +44 01904 321387.
1
Current address: Manchester Academic Health Science Centre, University of Manches
Postado por às

Nenhum comentário:

Postar um comentário

Assinar: Postar comentários (Atom)