I Nicolau1, D Ling, L Tian, C Lienhardt, M Pai. 1. Department of Epidemiology, Biostatistics and Occupational Health, McGill University, Montreal, Quebec, Canada.
Abstract
BACKGROUND: Systematic reviews are used to inform tuberculosis (TB) guidelines. However, there are no data on whether TB systematic reviews are conducted well and reported transparently. METHODS: We searched four databases for reviews published between 2005 and 2010. Methodological quality was evaluated using AMSTAR and quality of reporting was assessed using PRISMA. RESULTS: Of 152 articles, 137 (90%) met the inclusion criteria. Only 3 of 11 AMSTAR quality items were met in most reviews: appropriate methods to combine findings (67%), comprehensive literature search (72%) and presentation of characteristics of included studies (90%). The other eight items were met in 4-53% of the reviews. Only 4% of the reviews disclosed conflicts of interest. The majority of the PRISMA items were reported in more than 60-76% of the reviews. Only nine items were reported in less than 55% of the reviews, the lowest being the full-search strategy (30%), risk of bias across studies in the Methods (27%) and Results (21%) sections, and indication of a review protocol (15%). CONCLUSIONS: Systematic reviews in our survey were well reported but generally of moderate to low quality. Better training, use of reporting guidelines and registration of systematic reviews could improve the quality of TB reviews.
BACKGROUND: Systematic reviews are used to inform tuberculosis (TB) guidelines. However, there are no data on whether TB systematic reviews are conducted well and reported transparently. METHODS: We searched four databases for reviews published between 2005 and 2010. Methodological quality was evaluated using AMSTAR and quality of reporting was assessed using PRISMA. RESULTS: Of 152 articles, 137 (90%) met the inclusion criteria. Only 3 of 11 AMSTAR quality items were met in most reviews: appropriate methods to combine findings (67%), comprehensive literature search (72%) and presentation of characteristics of included studies (90%). The other eight items were met in 4-53% of the reviews. Only 4% of the reviews disclosed conflicts of interest. The majority of the PRISMA items were reported in more than 60-76% of the reviews. Only nine items were reported in less than 55% of the reviews, the lowest being the full-search strategy (30%), risk of bias across studies in the Methods (27%) and Results (21%) sections, and indication of a review protocol (15%). CONCLUSIONS: Systematic reviews in our survey were well reported but generally of moderate to low quality. Better training, use of reporting guidelines and registration of systematic reviews could improve the quality of TB reviews.