Contributions
Abstract: 55
Type: Scientific Session
Abstract Category: N/A
Introduction: The use of observational databases to assess the comparative effectiveness of drugs not compared in Randomized Controlled Trials (RCT) is becoming very popular in MS due to the availability of large databases collected in clinical practice. Statistical methods have been developed to correct for biases related to the lack of randomization, such as Propensity Score matching.
Objective: To show limitations of the comparative effectiveness in observational studies.
Methods: I will illustrate some examples taken from real MS datasets to show two main limitations of comparative effectiveness run on observational datasets:
1) the heterogeneity of outcome assessment across different centers or areas of the world, that can cause a strong bias when it is not possible to adjust for center effects;
2) an assessment of the residual bias in a situation when two non randomized groups are compared after adjusting for different methods based on Propensity Score methodology showing how these adjustments cannot be enough to completely account for the lack of randomization.
Results: Data showing a high degree of heterogeneity on EDSS progression in a cohort of RRMS patients according to geographical area will be presented. Data showing a Propensity Score matched comparisonof two placebo arms of two different clinical trials will be presentded, showing the superiority of a placebo over the other placebo even when all the possible adjustements for baseline unabalances were run.
Conclusions: RCT are very costly to be conducted, both in terms of time, resources, patients engagement, but they remain the only sperimental design that guarantees an unbiased way of assessing comparative efficay of drugs.
Disclosure: MPS received consulting fees from Biogen, TEVA, Novartis, Merck Serono, Genzyme, Roche, GeNeuro, Medday, Celgene, Actelion
Abstract: 55
Type: Scientific Session
Abstract Category: N/A
Introduction: The use of observational databases to assess the comparative effectiveness of drugs not compared in Randomized Controlled Trials (RCT) is becoming very popular in MS due to the availability of large databases collected in clinical practice. Statistical methods have been developed to correct for biases related to the lack of randomization, such as Propensity Score matching.
Objective: To show limitations of the comparative effectiveness in observational studies.
Methods: I will illustrate some examples taken from real MS datasets to show two main limitations of comparative effectiveness run on observational datasets:
1) the heterogeneity of outcome assessment across different centers or areas of the world, that can cause a strong bias when it is not possible to adjust for center effects;
2) an assessment of the residual bias in a situation when two non randomized groups are compared after adjusting for different methods based on Propensity Score methodology showing how these adjustments cannot be enough to completely account for the lack of randomization.
Results: Data showing a high degree of heterogeneity on EDSS progression in a cohort of RRMS patients according to geographical area will be presented. Data showing a Propensity Score matched comparisonof two placebo arms of two different clinical trials will be presentded, showing the superiority of a placebo over the other placebo even when all the possible adjustements for baseline unabalances were run.
Conclusions: RCT are very costly to be conducted, both in terms of time, resources, patients engagement, but they remain the only sperimental design that guarantees an unbiased way of assessing comparative efficay of drugs.
Disclosure: MPS received consulting fees from Biogen, TEVA, Novartis, Merck Serono, Genzyme, Roche, GeNeuro, Medday, Celgene, Actelion