Back to all discussions

Manager   •   about 2 years ago

IDEA/NEED: protocol burden scoring system

Clinical trials are not easy. The document that describes what needs to be done in a trial is called a protocol. Many are publicly available on ClinicalTrials.gov, if you select the filter for includes Protocol, and look for a link on the trial record. There is a specific section of protocols called an SoE or schedule of events (or schedule of activities) that describes, week by week, what procedures need to happen and when and how. This includes drug administration, taking blood samples, doing imaging scans like CT/MRI, invasive procedures like biopsies, and so forth.

The need of most clinical trial sponsors and designers is to understand and quantify the burden of these clinical trials, especially the burden to patients / participants.

One example of an approach to this that has been described publicly, and mathematically, is the Tufts CSDD work (Tufts Center for Study of Drug Development). Their method has weights for each procedure. In the simplified form, the method counts up the total number of procedures and multiplies the weights by each count, and outputs totals by procedure type and by overall protocol burden score. This can then be used to compare to benchmarks, to similar trials, to see if the burden is significantly higher or lower. There are also published equations estimating the potential impact on recruitment speed or drop-out rate. And there are more nuanced equations they have published about how things change by disease type, caregiver involvement, and other nuances of the trial.

Hackathon entrants can consider creating easy-to-use tools for study designers to help quantify and potentially reduce patient burden, in a very transparent and data-driven way. One additional element that it isn't always easy to extract the schedule of events -- particularly as there are often nuanced footnotes -- into a tabular form to then perform computations. So NLP technologies or large language models / LLMs like GPT may be useful at extracting the details of the protocol Schedule of Events into a structured form to then compute burden scores and options for burden reduction.

Here's a link to Zach Smith's paper from Tufts with all of the procedure-level weights and numbers:
https://link.springer.com/article/10.1007/s43441-021-00336-2

  • 2 comments

  •   •   about 2 years ago

    We’re pleased to see this project here! Patient burden assessments hold much potential in helping to inform protocol design decision making that will ultimately reduce the burden placed on clinical trial participants and improve clinical trial performance. Consistently in our work with sponsor companies, deciphering and counting the procedures in the SoE is the most time-consuming step. An assessment tool which can quickly extract the numbers from an SoE would go a long way toward making participant burden assessment possible on a larger scale that could be applied during the protocol planning and finalization stages.

  • Manager   •   about 2 years ago

    Thank you, Zak, for your valuable input!
    I'm eager to hear from others participating in this challenge about their perspectives on this crucial discussion topic of speeding up clinical trials. Does anyone have any ideas on how we can optimally analyze the procedures listed in a schedule of events? This list includes tasks a physician must perform during a patient visit, such as blood draws, MRIs, and tissue sampling. For reference, you can check out some examples at https://clinicaltrials.gov/ under study documents tick "study protocols"

Comments are closed.