DEMAND BETTER RESEARCH
UNIVERSAL BASIC STANDARDS ARE NECESSARY TO IMPROVE RESEARCH
DECLARATION TO IMPROVE BIOMEDICAL & HEALTH RESEARCH
It is extraordinary that we tolerate a situation in which most research activity is wasted. Research findings are mostly wrong and riddled with biases including inadequate or absent reporting and hidden conflicts of interest. Since Doug Altman blew the whistle on this scandal over 25 years ago there has been growing acceptance of the scale of the problem and the need for change. But specific actions must follow.
We are an international group of researchers and patients who believe that:
it is ethically untenable to remain complicit in the crises that undermine science,
there are simple measures which can improve the quality and openness, and
the public and patients have a right to full access of the research they fund
To start fixing science’s problems we need to rally around a specific set of achievable demands. The ideas we present are not new. Nor will they solve all the problems that face medical research. But they are simple measures which could bring rapid and significant improvements.
1. Mandatory registration of interests
Failure to disclose potential conflicts of interest is widespread. Conflicting interests do not necessarily invalidate research but their existence has to be known to interpret findings. Although voluntary registers of interest already exist uptake is low and the necessity of scouring many possible sources for voluntary disclosure is a barrier to transparency. There must be an expectation that everyone involved in research, from patient groups to researchers, from peer reviewers to funders, declare all their interests in a central and assessable location. This is eminently achievable. In the US, such an obligation already exists for physicians and there have been calls for the General Medical Council to require the same in the UK.
Over seven million researchers are already listed on the Open Research and Contributor ID (ORCID) system with registration already required by many institutions and journals. ORCID could catalogue declarations of interests with links embedded in publications. While the US experience suggests that legislation might be required to mandate disclosure, in the meantime, academic institutions, journals and funders could begin to work with ORCID to facilitate a comprehensive system of declaration.
2. That journals promote registered reports
Post-hoc abuses such as ‘outcome switching’ and so-called p-hacking are rife. The incentive to generate ‘interesting’ results drives undeclared repeated analyses to fit publishable narratives. Even when deviations from protocols are identified the supposed ‘self-correcting’ mechanism of peer scrutiny simply does not work. ‘Registered reports’ are an elegant solution. Authors submit their methods prior to data collection and analysis and if these satisfy peer review, journals commit to publication, regardless of the results generated. Authors may face greater workload in the planning of studies but benefit from peer review at the earliest stages. Registered reports differ from clinical trial registration and they are applicable to any hypothesis-driven research, rather than only clinical trials.
As of February 2020 only about 1% of medical journals allow authors to submit registered reports. We call on all journals to promote registered reports, with the intention that at least 20% studies in ICMJE journals are published as registered reports by 2025. We also call on public and charitable bodies to require that all hypothesis driven research they fund is published as registered reports within 10 years.
3. That all publicly funded research is registered and published in designated Research Repositories
The majority of research is funded by public and charitable funds. Yet, huge amounts of research is never published at all, which aside from being an indefensible waste of public money, is a major source of publication bias. Meanwhile, basic research documentation which is essential to ensure appropriate research conduct, such as protocols, are only sometimes available, either on voluntary databases or upon agreement of study authors.
The World Health Organization (WHO) has long urged registration of trials in affiliated ‘primary registries’, such as ClinicalTrials.gov and the EU Clinical Trials Register which can all be searched simultaneously through a dedicated WHO website. Mandatory registration of trials has improved transparency, although compliance with publication requirements is poor, possibly hampered by problems with the basic functionality of some major registries. Even where trials have been registered, usually only very limited information is shared, rather than the full protocols required to really understand study plans.
Most researchers don’t work in trials. Some principled scientists do register their work but while this remains voluntary such researchers are likely to remain in the minority. For all publicly funded research, not just trials, comprehensive documentation including protocols, statistical analysis plans, statistical analysis code and raw or appropriately de-identified summary data should be available on a single WHO affiliated repository, designated for that purpose by each state or groups of states.
Depositing documentation need not become onerous for researchers and could replace much of the overly bureaucratic reporting currently required for funders and ethics committees. Different solutions may exist in different countries. For example, England’s Health Research Authority could develop such a registry, by building on its existing public databases. Or, through additional national funding and international support, existing platforms which promote transparency and accessibility could be designated for this purpose through collaboration with national research bodies.
While basic solutions exist remaining complicity with the scandal of poor medical research is not defensible. The solutions we propose will not solve all of the problems, but we think they will be a start. Once they are achieved, further research will be vital to determine the effectiveness of these measures and to identify and remedy any unintended consequences.
We invite you to join the campaign to improve science. Please consider signing the declaration, or help us to develop the declaration by telling us how to improve it.