By Matt Fates, Partner at Innospark Ventures
At any given time, there are roughly 400,000 clinical trials occurring globally. In the US, there are ~85,000 clinical trials currently underway in the oncology space alone. Given the prevalence of common cancers in the US population and the sheer volume of trials, one would expect it would be a straightforward exercise to identify and enroll patients in these trials. Incentives could not be more aligned — Patients want better access to developing treatments, providers want to provide the best care possible, and pharmaceutical/biotech companies want to identify the efficacy of potential therapeutics as quickly as possible. Yet, this is not our reality today.
In fact, identifying and enrolling patients in clinical trials is one of the major barriers to bringing therapeutics to the market. The inability to expediently enroll and commence trials cannot be understated: prolonged trial starts delay research, burden budgets, and limit access to potentially life-saving treatments. According to the NIH, “globally, more than 80% of trials fail to enroll on time resulting in an extension of the study and or addition of new study sites,” both of which increase costs. Furthermore, “global data analysis of all terminated trials within Clinical Trials Databases reported 55% of trials were terminated due to the single highest reason of low patient accrual rate.” Again: this hurts pharmaceutical companies, drives up the cost of drugs and most importantly means many patients do not get access to potentially lifesaving or life prolonging medications in time.
In speaking with industry professionals, we learned that there is widespread recognition of this problem, but solutions are sparse today. One medical expert shared “queries are increasingly looking for needles in the haystack, because trial criteria have become more nuanced and specific. Searching for patients manually becomes complicated, especially because lots of valuable information is in the doctor’s handwritten notes. Even for large AMCs like MGH, which have in-house teams that do this, it is a very inefficient and time-consuming process.” Various patient searching and matching systems exist, but they are limited to high-level structured data fields, which falls far short of the complex criteria that can often only be found in patient notes or lab results. These existing systems can help narrow the search from 10,000s to 1,000s but then rely on medically trained professionals to do the individual reviews from there. This is very time consuming, prone to human error, and cannot practically be repeated regularly.
There are several reasons this dynamic persists in the clinical trial market, yet we believe that better use of data and AI can be a major solve to the problem. Dyania Health has built and brought to market an NLP-based expert system that leverages not only all the structured data within an electronic medical record (EMR), but also all the unstructured details in physician notes. This enables users to much more rapidly identify patients who match increasingly complex criteria, especially within oncology trials. Once more, it can be run daily to pick up any new diagnosis immediately, vs. waiting for a human reviewer to revisit the criteria checklist or expecting physicians to remember all the criteria for the various trials going on at their hospital (which can be 100s within large medical centers).
Dyania’s system is built on a deep understanding of the language, structure, and dependencies within medical records, medical ontologies and rules set up specifically by trained physicians, and paired with NLP to help parse and capture relevant language within the unstructured text. They have created a deep level of proprietary intelligence by extending well known medical taxonomies with clinical knowledge from a large number of expert clinicians. Endowed with this expert intelligence, the Dyania AI solution goes far beyond any other patient matching solution in the market, and is able to automatically identify and qualify patients more accurately, with a confidence score across the various inclusion and exclusion criteria for even the most complex clinical trials. This not only sources more relevant patient matches more quickly, more accurately and more regularly, but it also gives the principal investigator on site, and their staff, more time to then properly engage with the relevant doctor and their patients to determine if enrollment in the trial makes sense on an individual case by case level. The ultimate decision regarding enrollment and participation still resides with the patient and the medical professionals.
Dynaia’s approach to acquiring and embedding computational intelligence resonates with our view of an effective approach for intelligence acquisition for machines. In many problems, starting with a symbolic intelligence implementation and coupling it with intelligence from data often yields the most effective path to robust, deployable AI applications. Their intelligence is explainable and tractable and they can predictably expect to improve it with what they are able to learn algorithmically from data.
While the high-priority problem and well-architected AI-based solution are a compelling match, we are also excited by the prospect of backing and working closely with the founder and CEO, Eirini Schlosser. Eirini has an impressive personal backstory and is driven by a strong passion for solving this problem. The amount of knowledge and connections she has already accumulated will serve her well as she leads Dyania into the initial customer engagements and trials this year. Eirini has also surrounded herself with an impressive team of technologists, medical professionals and trusted advisors. Co-founder Dr. Dimitrios Iliopoulos (Harvard, UCLA) is a drug discovery expert and widely recognized for his impressive research (130+ published articles) and commercial success. Other key advisors include Dr. John Chelico (CMIO at Common Spirit Health), Dr. Allan Pantuck (Co-founder of Kita Pharma and Executive Chair of UCLA IRB Committee), Dr. Alexandra Drakaki (UCLA Oncologist) and Konstantine Arkoudas (Amazon Alexa, Bloomberg).
We are thrilled to partner with Eirini and her team and lead this round of funding, and welcome a great group of co-investors in the Outsiders Fund, Wild Basin, Big Pi and Tau Ventures as we all work to support the adoption and expansion of Dyania’s solution, driving down clinical trial costs and accelerating the delivery of new therapies to patients.