Article type
Year
Abstract
Background: Model-based meta-analysis (MBMA) is a technique increasingly used in drug development for synthesising results from multiple studies, allowing pooling of information on treatment, dose-response and time-course characteristics, which are often non-linear. Network meta-analysis (NMA) is used following systematic review for simultaneously comparing effects of multiple treatments and is frequently employed in epidemiology. Recently, a framework for dose-response MBNMA has been proposed that draws strengths from both MBMA and NMA.
Methods: We expand the MBNMA framework to allow for non-linear modelling of multi-parameter time-course functions for comparative effectiveness. This methodology preserves randomisation by aggregating within-study relative effects and, by modelling consistency equations on time-course parameters, it allows for testing of inconsistency between direct and indirect evidence. Residual correlation between observations can be accounted for using a multivariate likelihood. We demonstrate our modelling framework using an illustrative dataset of 24 trials investigating 29 treatments for pain in osteoarthritis and propose a step-by-step approach for model selection.
Results: For our dataset, we report results from 10 different models that employ a range of different time-course functions and simplifying assumptions. An Emax function allowed for the greatest degree of flexibility, both in the time-course shape and in the specification of time-course parameters. Our final model had a posterior mean residual deviance of 291.4 (compared to 345 data points), indicating a good fit to the data.
Conclusions: Time-course MBNMA combines strengths from MBMA and NMA to allow inclusion of multiple study time points into analyses in a statistically robust manner. This has the potential to substantially improve precision within meta-analyses, making it easier for healthcare providers and guideline developers to make the most informed decision regarding treatment options for patients, as well as potentially reducing the need to conduct additional clinical trials during drug development.
Patient or healthcare consumer involvement: The development of our research has not involved patients as it has not been linked to any specific health condition. However, we strongly recommend that any future analyses using MBNMA would seek to do so.
Methods: We expand the MBNMA framework to allow for non-linear modelling of multi-parameter time-course functions for comparative effectiveness. This methodology preserves randomisation by aggregating within-study relative effects and, by modelling consistency equations on time-course parameters, it allows for testing of inconsistency between direct and indirect evidence. Residual correlation between observations can be accounted for using a multivariate likelihood. We demonstrate our modelling framework using an illustrative dataset of 24 trials investigating 29 treatments for pain in osteoarthritis and propose a step-by-step approach for model selection.
Results: For our dataset, we report results from 10 different models that employ a range of different time-course functions and simplifying assumptions. An Emax function allowed for the greatest degree of flexibility, both in the time-course shape and in the specification of time-course parameters. Our final model had a posterior mean residual deviance of 291.4 (compared to 345 data points), indicating a good fit to the data.
Conclusions: Time-course MBNMA combines strengths from MBMA and NMA to allow inclusion of multiple study time points into analyses in a statistically robust manner. This has the potential to substantially improve precision within meta-analyses, making it easier for healthcare providers and guideline developers to make the most informed decision regarding treatment options for patients, as well as potentially reducing the need to conduct additional clinical trials during drug development.
Patient or healthcare consumer involvement: The development of our research has not involved patients as it has not been linked to any specific health condition. However, we strongly recommend that any future analyses using MBNMA would seek to do so.