Skip to main content

Languages

The value of author contact in determining risk of bias in included studies in a systematic review

Date and Location

Session: 

O2.05.2

Date

Saturday 21 September 2013 - 13:30 - 15:00

Location

Presenting author and contact person

Presenting author

Helen Atherton

Contact person

Helen Atherton
Abstract text
Background: Cochrane review methodology suggests that review authors consider contacting authors of included studies to obtain further information where risk of bias cannot be adequately assessed using published reports. This approach is not usually recommended in non-Cochrane reviews. It is unknown what difference author contact makes to this assessment, and whether this may affect the conclusions of a review. Objectives: Compare difference in risk of bias rating when rating is made based solely on published report versus published report and author contact. Methods: During the conduct of a Cochrane review on email for communication between healthcare professionals and patients, authors kept a record of study author contacts in relation to assessment risk of bias and whether clarification was obtained. This allowed for comparison between the risk of bias rating for each study before and after author contact. Results: Author contact was necessary in all nine included studies, as elements of study design were unclear/not described in all. Before author contact, 34 of 54 domains were unclear. Allocation concealment was unclear in all studies, random sequence generation in five, blinding in seven, incomplete outcome data in four and selective reporting in two. After author contact was completed 3/54 domains remained classified as ‘unclear,’ because three authors could not provide all detail requested. Eight authors were contactable via email; one author, of a study published in 1995, was only contactable via postal-mail. Conclusions: In our review, author contact reduced uncertainty when assessing risk of bias and was successful even for a study published more than 17 years ago. The conclusions of the review, that low quality of the included studies meant that effects of email could not be adequately assessed, was further supported by clarification of risk of bias status. We suggest further studies to establish the added value of this activity.