Industry-funded meta-analyses more likely to have favorable conclusions
Click Here to Manage Email Alerts
Industry-funded meta-analyses published in oncology journals reported favorable conclusions more frequently than those with no industry funding, according to findings published in JAMA Network Open.
The analysis included 93 meta-analyses published from 2018 through 2021 on five oncology journal websites. Researchers examined study characteristics and results, as well as information about study authors.
Investigators coded study conclusions as negative, positive or equivocal. They also coded each article's subject matter based on whether it may affect company marketing or profits.
Twenty-one studies had author funding from industry, and 17 (81%) of those reported favorable conclusions.
Nine studies received industry funding, and seven (77.8%) reported favorable conclusions.
Sixty-three studies had no study or author funding from industry, and 30 (47.6%) of those reported favorable conclusions.
“The findings from our study point to the larger issue of conflict of interest in scientific research and reporting. There are many advocates for greater transparency and disclosures as a solution,” study author Alyson Haslam, PhD, research data analyst in the department of epidemiology and biostatistics at UCSF School of Medicine, told Healio. “Yet, in our study, many of the meta-analyses did disclose funding, and there were still differences in positive findings between studies receiving industry funding and those that did not, suggesting that disclosure is inadequate in reducing or eliminating bias. In a perfect world, research would be separate from industry funding.”
Healio spoke with Haslam about why she and her colleagues conducted this study, the implications of the findings, and how the design and execution of meta-analyses could be improved to reduce bias.
Healio: What motivated you to conduct this review?
Haslam: Our motivation was simply to see what topics are being studied, who is conducting and funding the research, and the methodologies used in these types of studies. When reviews of meta-analyses are done, they often focus on a specific topic or question. We wanted to provide a high-level perspective of meta-analyses in oncology, and to determine if patterns exist. Are authors more likely to be from a certain geographical region, are they on the topic of drugs or some other topic, or are the findings often supportive or unsupportive of the topic being studied?
Healio: Can you describe your methodology?
We focused our analysis on studies published in top journals between 2018 and 2021, as these journals should have high standards for the quality of research they publish. Because we wanted to look at broad patterns, we had very few inclusion or exclusion criteria. Our results are largely descriptive, but we did want to see what factors were associated with positive study findings (ie, supportive of the topic being studied).
Healio: What did you find?
Haslam: Meta-analyses in top journals reported on a variety of tumor types, as opposed to focusing on the most common tumor types that people have. About 40% of meta-analyses used randomized data only — which is good for determining efficacy — and about 30% used observational data only, which is good for characterizing the natural history of a condition. About 40% of analyses had a last author who had published more than 10 meta-analyses, suggesting that many of these studies were written by people who were experienced at conducting these types of analyses. The main take-away is that having a first author with more prior meta-analyses publications and having study or author funding from industry were more likely to result in favorable study outcomes.
Healio: Why do you think industry-funded studies more often had positive conclusions?
There could be several reasons. First, industry is primarily concerned about the economic implications of the study results, which means there is a vested interest in funding avenues of research that will be economically advantageous. Industry-funded studies often use medical writers who have skills in how to advantageously present data. Second, unlike interventional clinical trials that have very specific primary outcomes, study methodologies and statistical plans, there is greater analytic flexibility for meta-analytic research. This can lead to the selection of statistical tools that produce more favorable outcomes, a focus on secondary outcomes (eg, surrogate outcomes) and omission/suppression of data that may not be supported by the views of the industry funder.
Healio: Do you plan to conduct more research in this area?
Haslam: Our lab is regularly evaluating the methodologies of studies so we can have a better understanding of biases in studies, and how to avoid bias in methodologies and reporting. Specific to writing/publication bias, we previously published that when medical writers are used, studies are more likely to report on surrogate outcomes, and often these studies have favorable outcomes.
We recently concluded an analysis looking at the effects of paid medical consulting companies in scientific writing. The studies written by these companies largely are supportive of the outcome — especially if the study evaluated a drug — but they also often advocate an unmet need, particularly if the topic of the study is descriptive of a condition.
Healio: What needs to change in terms of how meta-analyses are conducted?
Haslam: One solution could be to have independent third-party administrators accept industry funds for specific research questions, advertise requests for proposals, and match the funds to research groups with aligned interest and expertise. More realistically, reviewers and editors would catch instances of spin or outcome-switching — which are more likely to occur in industry-funded studies — during manuscript review. More broadly, there needs to be a change in culture where readers of scientific information are better able to critically detect bias.
References:
- Buck E, et al. JAMA Netw Open. 2023;doi:10.1001/jamanetworkopen.2022.54405.
- Haslam A, et al. JAMA Netw Open. 2023;doi:10.1001/jamanetworkopen.2023.18877.
For more information:
Alyson Haslam, PhD, can be reached at UCSF Mission Bay Campus, 550 16th St., Mission Hall: Global Health & Clinical Sciences Building, 2nd Floor, San Francisco, CA 94158; email: alyson.haslam@ucsf.edu.
Collapse
Read more about
Click Here to Manage Email Alerts