Investigating human influences in systematic reviews and meta-analyses to improve the reliability and validity of evidence syntheses

点击翻译按钮获取中文摘要

Billions of pounds are spent each year on health care treatments with the aim of improving and saving lives. But not all clinical research is published and what is published not be reliable. By bringing together similar studies, assessing their quality, and combined effects, more reliable findings can be used to guide future clinical and policy decisions. The gold standard of this method of reviewing studies is called systematic review. These large reports summarise all relevant studies in a clear and fair way to understand the benefits and drawbacks of new treatments. They are highly cited articles and regarded as objective and transparent research designs that are necessary to inform patient care.There are increasing reports that systematic reviews are being published that do not live up to their high quality reputation. Systematic reviews are receiving attention for contributing to research waste, varying massively in their quality, being misleading and serving conflicted interests. When authors are asked by publishing journals to declare conflicts of interest, they relate to narrow criteria about funds received from industry within the last few years. Whilst clear guidelines for best practice in systematic reviews are available for authors, they do not provide guidance on review teams themselves. Reviews are sometimes carried out by people who have an interest in what the results are (bias). Or review teams might not take all the necessary efforts to find and present a full picture of the evidence needed for decision makers to understand (passivity). Members of the review team might have their own view on a treatment and this might shape the way the review is done. Review teams may not be sufficiently skilled or have the resources to carry out a high quality review (insufficient resources). Inflexible review methods may result in expensive review projects which do not reflect issues that matter to patients. Also, sometimes reviews are published without funding and authors' motivation for this is not questioned. The effects that human influence has on systematic reviews may be especially important when looking at health care problems that are complex.Considering the reputation of systematic reviews as fair and objective processes, poor conduct could be more problematic than in clinical research. Health care evidence might not be evaluated fairly and patients may suffer because the health service may not provide the best treatments as a result. Methods: To fully explore the ways in which the researchers in the team can influence systematic reviews I plan to: 1) Search for and review studies documenting problems with systematic reviews that can occur due to team influence. 2) Work with experts in research methods to highlight risks and solutions for systematic review teams, helping to ensure that the findings of my research impact on systematic reviews that are carried out in the future. 4) Produce a dedicated website and guidance in a clear and accessible format about the risks of human influence in systematic reviews to help people assess systematic reviews for themselves.5) Test guidance with researchers to refine and revise, ensuring that they are useful to people who do or use systematic reviews.What impact will this research have? This research will describe what makes an appropriate team and what types of team influence can cause problems to systematic reviews. The framework and guidance will help people who do or pay for research that involves a systematic review to understand what factors they need to consider when planning, funding or doing a systematic review. To influence the wider research agenda a user-friendly website that explains the goals, methods and results of the research will be created. This research will lead to better systematic reviews, to inform health care decision making for health services for patients and the public and to influence future health research.