Article type
Year
Abstract
Background: A systematic review and network meta-analysis of acupuncture and other physical therapies for the relief of chronic pain due to osteoarthritis of the knee was undertaken. There are several different physical therapies, variations of the same type of physical therapy, and combinations of therapies for osteoarthritis of the knee pain. The definition of treatment groups need to enable useful interpretation of the results for policy.
Objectives: The objective of this communication is to present the implications of different intervention definitions on the network meta-analyses in this review.
Methods: Given the multiple possible combinations of physical therapies for osteoarthritis of the knee, combinations of the main interventions were excluded from the analysis to make the project manageable. For one network analysis, 22 main interventions were defined and each one had the possibility of five adjunct treatments resulting in 110 possible treatment groups. For another network analysis, the five adjunct treatments were grouped into one group, leaving 22 interventions defined. For another, all physiotherapy interventions were combined into one group to reflect options available to general practitioners in the UK.
Results: For the detailed network there were either no significant differences between the treatment effects of the interventions with the same principal physical therapy or the face validity was questionable. With more interventions defined, there were fewer trials per treatment comparison in the model and there was a low power to detect treatment effect differences. Results for the grouped interventions network were influenced by the new network structure and additional interventions and trials being captured within the broader treatment definition.
Conclusions: In complex network meta-analyses decisions made about the definition of interventions are important. A too great an emphasis on differences between interventions results in a large network comprised of few trials per treatment comparison in the model, with low power to detect treatment effect differences.
Objectives: The objective of this communication is to present the implications of different intervention definitions on the network meta-analyses in this review.
Methods: Given the multiple possible combinations of physical therapies for osteoarthritis of the knee, combinations of the main interventions were excluded from the analysis to make the project manageable. For one network analysis, 22 main interventions were defined and each one had the possibility of five adjunct treatments resulting in 110 possible treatment groups. For another network analysis, the five adjunct treatments were grouped into one group, leaving 22 interventions defined. For another, all physiotherapy interventions were combined into one group to reflect options available to general practitioners in the UK.
Results: For the detailed network there were either no significant differences between the treatment effects of the interventions with the same principal physical therapy or the face validity was questionable. With more interventions defined, there were fewer trials per treatment comparison in the model and there was a low power to detect treatment effect differences. Results for the grouped interventions network were influenced by the new network structure and additional interventions and trials being captured within the broader treatment definition.
Conclusions: In complex network meta-analyses decisions made about the definition of interventions are important. A too great an emphasis on differences between interventions results in a large network comprised of few trials per treatment comparison in the model, with low power to detect treatment effect differences.