Epidemiology, reporting characteristics and PRISMA-NMA adherence of systematic reviews with network meta-analysis: a meta-research study

Article type
Authors
Bargeri S1, Castellini G1, Gianola S1, Guida S1, Lunny C2, Moja L3, Ravot G4
1 Unit of Clinical Epidemiology, IRCCS Istituto Ortopedico Galeazzi, Milan, Italy
2Knowledge Translation Program, Li Ka Shing Knowledge Institute, St. Michael's Hospital, Toronto, Canada; Cochrane Hypertension Review Group, the Therapeutics Initiative, University of British Columbia, Canada
3Department of Health Product Policy and Standards, World Health Organization, Geneva, Switzerland
4University of Milan, Milan, Italy
Abstract
Background: Network meta-analyses (NMA) have become increasingly popular as an innovative technique to compare multiple interventions within a network of studies.

Objective: To examine the epidemiology, reporting characteristics, and adherence to the PRISMA-NMA extension of a representative sample of published SRs with NMAs.

Methods: Cross-sectional study (protocol at https://osf.io/pa6dz/). We searched Pubmed for all NMAs indexed during January 2023. We extracted epidemiological (eg, journal, country, authors, NMA focus) and reporting (eg, number of studies, participants, nodes of interventions, outcomes, statistics, certainty of the evidence [CoE], PRISMA-NMA checklist, protocol registration) characteristics. We assessed the overall adherence to PRISMA-NMA. We explored the epidemiological and reporting characteristics that might affect quality of reporting and PRISMA-NMA adherence.

Results: A sample of 139 NMAs was included. Overall, 77% of NMAs were published on specialty journals with a median impact factor of 4 (interquartile range [IQR] 2.8–6.3) and had a median number of 7 authors (IQR 5–10), often based in China (52%). The majority of NMAs focused on therapeutics (92%), primarily targeting cardiovascular diseases (18%) (Table 1). The type of interventions addressed were mainly pharmacological (49%). On average, each NMA included 19 studies (median, IQR 12–40.5) with 3,415 participants (median, IQR 1,472–9,384). The majority (61%) prospectively registered a study protocol and reported three outcomes (median, IQR 2–5). Half (49%) reported having followed PRISMA-NMA, whereas 33% reported CoE (Table 2). The number of studies adherent to PRISMA-NMA shown a median of 72% with 35 of 49 items reported (IQR 33–38). The less reported items were limitations (6.5%, n = 9), funding (7.9%, n = 11), and study appraisal (21.6%, n = 30) in abstracts, the geometry of network (15.1%, n = 21) in methods, and results of individual studies (9.4%, n = 13) in results sections (Figure 1). NMAs with a registered protocol assessing CoE, published in a journal with IF > 10, were more likely to have higher adherence to the PRISMA-NMA (p < 0.05).

Conclusions: Most NMAs focused on pharmacological interventions are published in specialty journals with high impact factors, but their quality of reporting can be easily improved.