Conflicts Of Interest and Risk of Bias In Systematic Reviews On Interventions For 6 Common Diagnoses in Primary Care

Date & Time
Wednesday, September 6, 2023, 12:30 PM - 2:00 PM
Location Name
Session Type
Research integrity and fraud
Czajkowski M1, Snellman A2, Olsson L2
1Department of Medical Sciences, Örebro University, Sweden
2HTA-unit Centre for Assessment of Medical Technology in Örebro, Örebro University Hospital, Sweden

Background Knowledge on the association between disclosed conflicts of interest (COI) and risk of bias (RoB) in systematic reviews (SRs) is limited, especially from a clinical perspective. Objectives We aimed to explore the association between COI and RoB in a recent sample of SRs of clinical relevance. Methods A cross-sectional study of SRs on interventions for chronic obstructive pulmonary disease (COPD), diabetes mellitus (DM) type 2, hypertension, dementia, depression, and osteoarthritis published in 2019 was designed. PubMed was searched by a librarian. Two independent reviewers selected SRs using pre-specified criteria. Disclosures of COI and funding were extracted, and SRs were categorized as “exposed” (disclosing either financial COI or funding from industry) and “non-exposed” (disclosing no financial COI and no funding from industry). A random sample of 50 SRs was drawn from the exposed and the non-exposed category, respectively. Two independent researchers, blinded to the COI and funding disclosures, applied a ROBIS item for RoB assessment. The ROBIS tool is organized into four domains, and, if one of them is found associated with a high RoB, the others cannot make up for this. A working model based on “early stop” at the first domain implied that a high RoB was adopted. Results In total, 590 SRs were included; 189 (32%) evaluated medications, 161 (27%) focused on diabetes, and 22 (4%) were Cochrane reviews. COI were disclosed in 120 (20%) SRs, of which 100 (17%) implicated individual financial COI. Funding from industry was disclosed in 33 (6%) SRs. In all, 109 (18%) SRs were exposed, 362 (61%) were non-exposed to financial relationships, and 119 (20%) could not be categorized. High RoB was found in 43/51 (84%) exposed and 42/52 (81%) non-exposed SRs (p=0.64). The main reason was shortcomings of the second domain, i.e., identification and selection of primary studies in 24 (56%) and 29 (69%) SRs, respectively (p=0.21). Conclusions Most SRs in this clinically relevant sample were assessed as having high RoB, but it did not differ between SRs exposed and non-exposed to financial relationships. Further studies are needed to validate and clarify the implications of this finding. Patient involvement None.