Article type
Year
Abstract
Background: Systematic reviews (SRs) include all studies in a field, regardless of trial registration. Recent claims stated that unregistered trials are of lower quality than registered trials and their inclusion downgrades the quality of SRs. It was suggested that unregistered trials conducted post 2009 should be excluded from reviews, however there is no evidence to support this.
Objectives: To investigate how prevalent registration is for published fertility trials and if registration is associated with a decreased risk of bias.
Methods: The Cochrane Gynaecology and Fertility Groups’ Specialised Register was searched for full text, English fertility trials published from 2010 to 2014. A computer-generated list randomly selected 25 registered and unregistered trials per year. These 250 trials were assessed for methodological quality using the Cochrane 'Risk of bias' (RoB) tool, judged as being at low or high/unclear RoB, and analysed using an odds ratio (OR) with 95% confidence intervals (CI).
Results: A total of 693 trials met the inclusion criteria, 45% of which were registered. For each year there were more unregistered than registered trials published in journals. Registered trials were more likely to have a low RoB for random sequence generation (OR 2.80, 95% CI 1.60 to 4.90), allocation concealment (OR 2.38, 95% CI 1.39 to 4.01) and reporting the planned primary outcome from the protocol (OR 61.98, 95% CI 21.39 to 179.55). There was no difference between the RoB for registered and unregistered trials for blinding, incomplete outcome data or non-reporting of patient-centred outcomes (eg. live birth) (OR 1.36, 95% CI 0.81 to 2.27).
Conclusions: Registered trials were more likely to be considered to be at low risk in the categories of random sequence generation, allocation concealment and selective reporting. However as only 45% of fertility trials were registered, the exclusion of unregistered trials from SRs would greatly reduce the number of trials included, potentially introducing publication bias and reducing the power.
Objectives: To investigate how prevalent registration is for published fertility trials and if registration is associated with a decreased risk of bias.
Methods: The Cochrane Gynaecology and Fertility Groups’ Specialised Register was searched for full text, English fertility trials published from 2010 to 2014. A computer-generated list randomly selected 25 registered and unregistered trials per year. These 250 trials were assessed for methodological quality using the Cochrane 'Risk of bias' (RoB) tool, judged as being at low or high/unclear RoB, and analysed using an odds ratio (OR) with 95% confidence intervals (CI).
Results: A total of 693 trials met the inclusion criteria, 45% of which were registered. For each year there were more unregistered than registered trials published in journals. Registered trials were more likely to have a low RoB for random sequence generation (OR 2.80, 95% CI 1.60 to 4.90), allocation concealment (OR 2.38, 95% CI 1.39 to 4.01) and reporting the planned primary outcome from the protocol (OR 61.98, 95% CI 21.39 to 179.55). There was no difference between the RoB for registered and unregistered trials for blinding, incomplete outcome data or non-reporting of patient-centred outcomes (eg. live birth) (OR 1.36, 95% CI 0.81 to 2.27).
Conclusions: Registered trials were more likely to be considered to be at low risk in the categories of random sequence generation, allocation concealment and selective reporting. However as only 45% of fertility trials were registered, the exclusion of unregistered trials from SRs would greatly reduce the number of trials included, potentially introducing publication bias and reducing the power.