How to Improve Clinical Trials in the US

Below are some short comments that the Good Science Project filed on two White House requests for information on clinical trials. We are optimistic that improving clinical trial infrastructure in the US would vastly improve our lifespan and quality of life.

*******

These comments are filed by Stuart Buck and Betsy Ogburn (a bioscientist at Johns Hopkins) on behalf of the Good Science Project (a think tank) in response to two White House Requests for Information: 

Clinical Research Infrastructure and Emergency Clinical Trials, 87 Fed. Reg. 64821, and,

Data Collection for Emergency Clinical Trials and Interoperability Pilot, 87 Fed. Reg. 65259.

            As the White House has correctly noted, the current clinical trial infrastructure in the United States is “not well prepared” to perform large and rigorous clinical trials in the case of emergencies or pandemics. One particular instance of this lack of preparedness involves the design of trial protocols and the capture of data in an efficient and interoperable format. 

            At the Good Science Project, we believe that the White House has identified real problems. Nonetheless, the narrow focus on emergencies and pandemics is misguided. A robust clinical trial infrastructure (including data standards) cannot be a mostly-theoretical plan that is left on the shelf, to be used only in case of emergency or only for specific diseases. The only way that such infrastructure will be usable at all is if it is constructed and used on an ongoing basis. In that case, it could be readily redeployed in case of emergency; but if not otherwise in use, we would still be starting essentially from scratch in future emergencies.   

I.          The Goal Should Be To Create an Efficient Clinical Trial Infrastructure Period, Not Just For Emergencies

We should create an efficient clinical infrastructure to be used on an ongoing basis, not just in so-called “emergencies.” This is true for multiple reasons. 

For one thing, drawing a line between emergency and non-emergency settings is arbitrary. Are leading causes of death and morbidity–-like cancer, heart disease, Alzheimers, and mental health disorders–-less worthy of attention than pandemics? They may get less attention from journalists or politicians, but they remain leading causes of death and disability–-year in, year out.

For another thing, drawing a line between emergency and non-emergency settings will be counterproductive and will only lead to failure. Clinical trial infrastructure is not like a hatchet kept behind glass, to be broken only in case of emergencies. Instead, clinical trial infrastructure needs to be built out and deployed to be sure that it works at all, and then we need to iterate and improve on an ongoing basis. If left for so-called “emergencies,” chances are good that any infrastructure or coordination protocols would fail, if only because no one had never worked out the kinks.

Thus, we recommend building systems for protocol harmonization, data sharing, and encouraging (or even engineering) large multi-site trials, and deploying them now for all important clinical questions. Not only will such efforts improve clinical research across the board, such an infrastructure would be ready to redeploy in case of a fast-paced emergency like a pandemic. Perhaps in “emergency” settings, one might want to force institutions and PIs to waive the right to first publication(s), but otherwise the infrastructure and data standards should be the same. 

II.        How To Improve Our Ability to Accumulate Evidence From Clinical Trials

A related problem with the current clinical trial infrastructure is that we are failing to fund high-quality trials that are coordinated in advance, as well as to properly pool and accumulate evidence.

First, we need to build infrastructure to tap into non-academic medical centers with clinical trial infrastructure. Pharma companies routinely contract with small medical centers around the country (including in rural areas) to run clinical trials. These centers have excellent trial infrastructure and, anecdotally, are often interested in academic research but have no entry point in most cases. NIH, perhaps through the CTSAs, and/or PCORI should develop a network of these sites that could be deployed to increase the sample size and generalizability of high-priority funded trials, and to have at the ready for “emergencies.”

Second, ClinicalTrials.gov is not a sufficient data-sharing platform for federally funded research. Data that was paid for by the US government must be made maximally useful for improving public health and clinical practice. An idea from Barbara Bierer (and others) is that we should incentivize data “authorship” to give credit for producing data that is used for productive research down the road. For example, NIH, PCORI, and others could prioritize the creation of useful data over publication of first-author papers when evaluating PIs for future funding, because the production of data is what most advances our collective knowledge base.

Third, NIH could consider moving towards a contract, rather than grant, model of funding for some high-priority clinical questions. This is similar to PCORI’s model, and would better enable NIH to make decisions on behalf of PIs, such as requiring multiple teams to come together in a multisite trial, enforcing the adoption of a common protocol, expanding a trial’s footprint by enlisting non-academic sites to enroll more patients, etc.

Fourth, NIH should try to completely eliminate “small crappy trials.” While there are many pilot and feasibility trials that may be worth funding, there are still too many “small crappy trials” that get funded for some reason. NIH should make a considered effort to fund fewer and better trials along the lines of ACTIV, CAST, WHI, NLST, and more. Single-arm trials (with no randomization) should not be funded except in the most extenuating of circumstances. And in many cases, this may require lifting the NIH’s de facto $500,000 limit on direct costs per year. 

            Finally, when it comes to data systems and standards, we should aim to 1) include at least the possibility of randomization at the point of care, so that many more doctors and health care facilities are enabled to participate in clinical trials with a lower barrier of entry; and even more ambitiously, 2) rebuild the electronic health care records systems in the US to tailor them for research rather than merely for reimbursement.

The latter would make clinical trials far less expensive to run and would drastically lower barriers to pooling evidence across institutions. The Office of the National Coordinator (ONC) should make that their top aim and priority when it comes to interoperability.