Quality and quantity of cancer-related systematic reviews published in high-impact journals

Tags: Oral
Goldkuhle M1, Dahm P2, Narayan V3, Skoetz N1
1Cochrane Cancer Alliance, University Hospital of Cologne, Department I of Internal Medicine, Cologne, Germany, 2Cochrane Cancer Alliance, Minneapolis Veterans Affairs Health Care System and University of Minnesota, Department of Urology, USA, 3Minneapolis Veterans Affairs Health Care System and University of Minnesota, Department of Urology, USA

Background: Systematic reviews (SRs) play a critical role in guiding evidence-based clinical practice including the management of patients suffering from cancer. Cochrane is recognized for its contributions to the development of SR methodology and its dissemination, which has contributed to publication of SRs in many other journals.

Objectives: To assess the scope and quality of SRs published in high-impact medical journals.

Methods: Following a written a priori protocol we performed a comprehensive search for SRs in PubMed published in high-impact general medical journals (e.g. NEJM, Lancet, BMJ etc.) and leading cancer journals (e.g. JNCI, JCO, Lancet Oncology etc.) over a five-year period (2011-2016). Two review authors performed all steps of the review independently in duplicate. We used AMSTAR (A Measurement Tool to Assess Systematic Reviews) to assess methodological quality of the SRs.

Results: We identified 221 SRs that met our inclusion criteria: most of these were intervention reviews, 36 SRs without meta-analysis (MA), 41 including individual patient data, 15 evaluating prognostic factors or models, seven assessing diagnostic test accuracy, six network meta-analyses and one overview of reviews. Sixty-nine intervention reviews with MA were based on randomized controlled trials (RCTs), 93 on observational data. Rating of SRs with a MA based on RCTs shows that the most reported topic is cancer in general, especially adverse events of drugs. The average number of RCTs was 24 and the average number of participants 8411. Quality indicating items such as the number of abstractors and databases used are often satisfactory, whereas serious lacks occur in fields like a priori design (20%) and assessment of publication bias (46%). The quality of included studies is rarely evaluated in sensitivity analyses (29%).

Conclusions: A growing number of cancer-related reviews are published in high impact journals. These are of variable quality, with notable shortcoming in the area of a priori design, evaluation of publication bias and integration of quality aspects in analyses. There continues to be an need to raise the quality of cancer SRs.