Article type
Year
Abstract
Background:
Non-randomized comparative studies can be a source of effectiveness data, particularly in the absence of randomized studies. However, given the risk of selection bias, it seems reasonable to expect such studies to make some attempt at comparability i.e. similar populations allocated to each treatment. One would not expect treatments to have been explicitly selected based on patient prognosis.
Objective:
To assess the extent to which non-randomized comparative studies are subject to explicit selection on the basis of prognosis, using a systematic review of fenestrated or branched endovascular abdominal aortic aneurysm (fEVAR or bEVAR).
Methods:
A systematic review was conducted following published guidance. Bibliographic databases such as Medline (OvidSP), EMBASE and CENTRAL were searched on 1 October 2013 without date limits. To be included, non-randomized comparative studies had to demonstrate no explicit intent to select the treatment based on prognosis.
Results:
Twenty-four studies were ordered for full text screening, but none satisfied our inclusion criteria. Nineteen of these studies were excluded due to study design; five of these were found to have selected treatment on the basis of prognosis despite the fact that they were described as comparative studies. Therefore, these studies were considered to have selection bias and thus non-comparable groups. Moreover, the studies had included patients in the fEVAR/bEVAR group who were actually ineligible for the comparator, open surgical repair (OSR) due to, for example old age (Table 1).
Conclusion:
In fEVAR/bEVAR not one non-randomized study attempted comparability of patients. Indeed, they explicitly used the same method of selection i.e. according to prognosis. Indeed, some patients treated with fEVAR/bEVAR were not even eligible for OSR. This is only one example and it seems unlikely that it is unique. We recommend research to investigate comparability in other clinical areas, particularly where randomized trial evidence is scarce. Researchers and publishers should take particular care when describing or assessing what can be informed from this type of study design.
Non-randomized comparative studies can be a source of effectiveness data, particularly in the absence of randomized studies. However, given the risk of selection bias, it seems reasonable to expect such studies to make some attempt at comparability i.e. similar populations allocated to each treatment. One would not expect treatments to have been explicitly selected based on patient prognosis.
Objective:
To assess the extent to which non-randomized comparative studies are subject to explicit selection on the basis of prognosis, using a systematic review of fenestrated or branched endovascular abdominal aortic aneurysm (fEVAR or bEVAR).
Methods:
A systematic review was conducted following published guidance. Bibliographic databases such as Medline (OvidSP), EMBASE and CENTRAL were searched on 1 October 2013 without date limits. To be included, non-randomized comparative studies had to demonstrate no explicit intent to select the treatment based on prognosis.
Results:
Twenty-four studies were ordered for full text screening, but none satisfied our inclusion criteria. Nineteen of these studies were excluded due to study design; five of these were found to have selected treatment on the basis of prognosis despite the fact that they were described as comparative studies. Therefore, these studies were considered to have selection bias and thus non-comparable groups. Moreover, the studies had included patients in the fEVAR/bEVAR group who were actually ineligible for the comparator, open surgical repair (OSR) due to, for example old age (Table 1).
Conclusion:
In fEVAR/bEVAR not one non-randomized study attempted comparability of patients. Indeed, they explicitly used the same method of selection i.e. according to prognosis. Indeed, some patients treated with fEVAR/bEVAR were not even eligible for OSR. This is only one example and it seems unlikely that it is unique. We recommend research to investigate comparability in other clinical areas, particularly where randomized trial evidence is scarce. Researchers and publishers should take particular care when describing or assessing what can be informed from this type of study design.