A comparative study of publication bias in systematic reviews of traditional Chinese medicine and conventional medicine

Tags: Poster
Wang D1, Hu R1, Wen L1, Fang S1, Zhao N1, Tan L1, Lu C1, Liu X1, Wang X1, Zhang Y1, li J1, Li X1, Liu J1, Song F2, Fei Y1
1Beijing University of Chinese Medicine, 2University of East Anglia

Background:

With the development of the methods to detect and reduce publication bias, the impact of publication bias on the results of systematic reviews (SR) should be reduced. As global interest in traditional medicine grows, the number of published SRs of traditional Chinese medicine (TCM) has increased greatly during the last twenty years.

Objectives:

We aim to compare the measures taken to prevent and detect publication bias in TCM or non-TCM SRs.

Methods:

We searched PubMed and Wanfang (a Chinese database) for TCM or non-TCM SRs published from 2001 to 2004, 2005 to 2007, and in 2010 and 2014. We selected the retrieved SRs randomly using SAS software; a total of 160 TCM SRs and 160 non-TCM SRs in English from PubMed, and 160 TCM SRs and 160 non-TCM SRs in Chinese from Wanfang were selected. The included SRs were screened by two people independently who extracted details of the basic characteristics of the SRs, methods used to prevent and detect publication bias, and methods to reduce its impact.

Results:

The percentage of SRs that reported methods to detect publication bias were 63% in the year 2001, and 95% in 2014; 87% in SRs in the Chinese language and 93% in English; 95% in non-TCM SRs, and 96% in TCM SRs. In fact, 98% SRs may have a problem of publication bias . Sixty-nine per cent of the reviews used funnel plots or related methods to test for publication bias.

Conclusions:

Recent reviews showed significant improvements in the measures taken to prevent and detect publication bias. Methods used to detect publication bias in English language SRs were more clearly reported than that in Chinese language SRs; TCM SRs were largely the same as non-TCM SRs with regard to the quality of methods used for publication bias. At present, the funnel plot is one of the most commonly used methods.

Patient or healthcare consumer involvement:

The users of this study will be authors of SRs. All the authors write SRs, and include clinicians and methodologists.