Article type
Year
Abstract
Introduction:
Objectives:
Methods: A meta-analysis of behavioural therapy for chronic low back pain has been conducted within the framework of the Cochrane Collaboration. Twenty RCTs were included in this review. Within every study, standardised mean differences (SMD) were computed for each outcome variable so that a positive SMD indicated a positive effect. In our review protocol we had defined six domains for outcomes that we considered the most important, i.e., behavioural outcomes, overall improvement, back specific functional status, generic functional status, return to work, and pain intensity. Most studies used a variety of outcomes within the behavioural domain, such as pain behaviour, anxiety, depression, coping, and cognitive errors, to assess the effect of behavioural therapy. If studies included multiple outcome variables, we computed a mean SMD for that domain. Separate meta-analyses were performed for post-treatment and long-term results for the various domains using the random effects model.
Results: Of the 11 studies comparing some type of behavioural treatment with no treatment or waiting list controls, 7 reported data mat could be used in the statistical pooling. The weighted SMD (WSMD) for the behavioural domain post-treatment was 0.40 (95% CI 0.10 ; 0.70). Three of the four studies that were not included in the pooling showed similar small positive effects. Six studies compared behavioural treatment in addition to another treatment with that other treatment alone. The WSMD for the behavioural domain was 0.19 (95% CI -0.08 ; 0.45) post-treatment and for the long-term follow-up 0.32 (95% CI -0.06 ; 0.71) .
Discussion: We would like to discuss several problems that we encountered in our meta-analysis. First, post-treatment and long-term follow-up results of each outcome variable within each domain were used to compute SMDs because change-scores were not reported in the original studies. SMDs ignore baseline dissimilarities among groups, which may occur in small trials, when they are based on post-treatment results. Second, mean SMDs provide a global idea about the effectiveness of behavioural treatment but ignore the fact that for some behavioural outcomes there may be a positive effect and for other outcomes a negative effect. Third, when authors of a study have used multiple outcomes for a specific domain, they may only have reported the statistically significant outcomes in the final article. Computing a mean SMD per domain for each study using the reported data may result in an overestimation of the real treatment effect. Interpretation of mean SMDs per study is difficult and translation into clinical practice not easy.
Objectives:
Methods: A meta-analysis of behavioural therapy for chronic low back pain has been conducted within the framework of the Cochrane Collaboration. Twenty RCTs were included in this review. Within every study, standardised mean differences (SMD) were computed for each outcome variable so that a positive SMD indicated a positive effect. In our review protocol we had defined six domains for outcomes that we considered the most important, i.e., behavioural outcomes, overall improvement, back specific functional status, generic functional status, return to work, and pain intensity. Most studies used a variety of outcomes within the behavioural domain, such as pain behaviour, anxiety, depression, coping, and cognitive errors, to assess the effect of behavioural therapy. If studies included multiple outcome variables, we computed a mean SMD for that domain. Separate meta-analyses were performed for post-treatment and long-term results for the various domains using the random effects model.
Results: Of the 11 studies comparing some type of behavioural treatment with no treatment or waiting list controls, 7 reported data mat could be used in the statistical pooling. The weighted SMD (WSMD) for the behavioural domain post-treatment was 0.40 (95% CI 0.10 ; 0.70). Three of the four studies that were not included in the pooling showed similar small positive effects. Six studies compared behavioural treatment in addition to another treatment with that other treatment alone. The WSMD for the behavioural domain was 0.19 (95% CI -0.08 ; 0.45) post-treatment and for the long-term follow-up 0.32 (95% CI -0.06 ; 0.71) .
Discussion: We would like to discuss several problems that we encountered in our meta-analysis. First, post-treatment and long-term follow-up results of each outcome variable within each domain were used to compute SMDs because change-scores were not reported in the original studies. SMDs ignore baseline dissimilarities among groups, which may occur in small trials, when they are based on post-treatment results. Second, mean SMDs provide a global idea about the effectiveness of behavioural treatment but ignore the fact that for some behavioural outcomes there may be a positive effect and for other outcomes a negative effect. Third, when authors of a study have used multiple outcomes for a specific domain, they may only have reported the statistically significant outcomes in the final article. Computing a mean SMD per domain for each study using the reported data may result in an overestimation of the real treatment effect. Interpretation of mean SMDs per study is difficult and translation into clinical practice not easy.