Types of selective inclusion and reporting bias in randomised trials and systematic reviews of randomised trials

Article type
Authors
Page M1, McKenzie J1, Green S1, Forbes A1
1School of Public Health and Preventive Medicine, Monash University, Melbourne, Australia
Abstract
Background: There are many ways in which outcome data can be selectively included or reported in randomised trials and systematic reviews (e.g. including the most favourable effect estimate in a meta-analysis when trialists report data for multiple scales). To date, no comprehensive list of the types of selective inclusion and reporting bias in trials and reviews has been collated.

Objectives: To increase trialists’ and review authors’ awareness of scenarios where they may inadvertently introduce bias through selective inclusion or reporting of results, by collating a list of these practices.

Methods: Searches of the Cochrane Methodology Register, PubMed, and PsycInfo were conducted in April 2011. Methodological papers reporting types of selective inclusion or reporting bias were included. Text describing examples of selective inclusion or reporting bias was extracted and assigned to categories (some pre-specified, others developed iteratively).

Results: Two-hundred and eighty-nine publications were included. Most reported solely on selective reporting in trials (n = 225; 77.9%); 16 (5.5%) reported solely on selective inclusion or reporting in systematic reviews; and 48 (16.6%) reported on both. Eight categories were defined: complete omission of outcomes; selective inclusion or reporting of specific measurements of outcomes; selective inclusion or reporting of specific analyses of outcome data; selective inclusion or reporting of subgroups of outcome data; partialreporting of outcomes; selective modification of outcomes across different sources or sections of a publication; selective modification of analyses across different sources or sections of a publication; and selective reporting of meta-analytic effect estimates. Forty-six subcategories were defined.

Conclusions: An extensive list of types of selective inclusion and reporting bias that can occur in trials and reviews was collated. Increasing trialists’ and review authors’ awareness of these practices is a first step to minimising their occurrence.