Background: Language criteria for primary studies vary between
systematic reviews: some have no restrictions, whereas others consider
only publications in English or in English and other specified languages. As
translations require considerable resources, these restrictions are understandable;
however, they may cause language bias. Objectives: To assess
the validity of a resources-saving selection procedure for non-English
studies in systematic reviews. Methods: For a systematic review on
antidepressants, two reviewers assessed the relevance of identified non-
English, i.e. Chinese studies, on basis of the English abstract. These studies
were further evaluated only if the abstract clearly indicated fulfilment of
the study inclusion criteria (‘relevant’); they were excluded if the abstract
contained study exclusion criteria (‘not relevant’) or insufficient information
on study inclusion criteria to allow evaluation (‘unclear’). In contrast,
‘unclear’ English studies underwent an additional full text review. For the
analysis of the validity of this study selection procedure, the first two
reviewers and two further reviewers blinded to the initial results classified
Chinese studies identified in the literature search on basis of the translated
full text. For studies classified as ‘unclear’ on the abstract level and
excluded, the agreement regarding study exclusion between the abstract
level and full text level was compared. The full texts of studies classified as
‘relevant’ were also assessed as controls. Results: Of the 14 Chinese
citations identified, all had English abstracts. On basis of these abstracts,
the first two reviewers classified two studies as ‘relevant’, eight as ‘unclear’
(with subsequent exclusion), and four as ‘not relevant’. These reviewers
and the two additional reviewers found no disagreement in study inclusion
or exclusion between the abstract level and full text level for studies
initially classified as ‘unclear’. Of the two studies initially assessed as
‘relevant’, one was excluded on the full text level. Results for other
languages will be presented. Conclusions: Our preliminary findings
suggest that the exclusion of non-English studies on basis of the English
abstract is a valid indicator of their true lack of relevance. Thus this might
be a resources-saving selection procedure to include relevant non-English
studies in systematic reviews.