Why Clinical Trials Need Digital Endpoints
Geoffrey Gill, MS, Founder and CEO, Verisense Health
The need to substantially improve clinical trials has never been more urgent. Costs are increasing, pricing pressure is growing, and the return on investment (ROI) on development is shrinking. Most approaches to improvement do not address the primary challenge that many outcome measures are fundamentally unreliable. Digital endpoints can address this issue and transform clinical trials.
According Citeline, the average success rate for bringing a Phase 1 candidate to market has dropped to an all-time low of just 6.7% and the probability of success for a Phase 3 trial dropped from 55% to 45% from 2006 to 2023. At the same time, there is increasing pressure for the United States to regulate drug prices with the Inflation Reduction Act and multiple proposed bills in Congress. Even without the impact of these regulations, the average peak sales of a new drug fell by 30% from 2013 to 2023.
Pharmaceutical companies and clinical research organizations (CROs) have recognized these challenges and are implementing a wide variety of programs to improve efficiency to improve efficiency" to cover key points missing from the sidebar: "They include eConsent, electronic enrollment methods, document and protocol automation, and decentralized clinical trials They are bringing benefits, but the complexity of clinical trials is growing faster than the benefits of these programs are being realized. According to Deloitte, from 2013 to 2023, the average cost of a clinical trial increased by 76% and the overall ROI of pharmaceutical innovation dropped 37% to 4.1% – less than the three-month U.S. Treasury Bill rate at the time of this writing.
Even more important than the financial performance of clinical trials are the time and other impacts. According to research from the Tufts Center for the Study of Drug Development (CSDD), U.S. Food and Drug Administration (FDA-) approved drugs and biologics spent an average of 89.8 months in clinical trials between 2014 and 2018, compared to 83.1 months between 2008 and 2013. This means patients are waiting longer to get new treatments.
Furthermore, the increased time and complexity are putting tremendous stress on the entire system – particularly at clinical trial sites. Go to any conference on clinical trials and a major, if not the dominant, theme is capacity issues at clinical sites and the difficulty recruiting them. The entire clinical trial infrastructure and process are close to breaking point.
This is not at all to say that these improvement efforts are wasted and should be stopped. They all address real issues and are useful. However, it must be recognized that they are ultimately efficiency improvement efforts and do not address the underlying causes of the growing size and complexity of clinical trials. Without addressing these causes, any improvement in the performance of clinical trials will be limited.
Fundamental Challenge of Clinical Trials
There are two major drivers for the size of any statistical study: the effect size of the phenomenon being studied and the reliability of the outcome measurement. The size of the treatment effect is determined by the drug itself, not the trial. Thus, the fundamental challenge with clinical trials is the unreliability of the outcome measures. There are two major issues with most of our current measures; they are only measured periodically and/or they are subjective.
Almost any physical measure is subject to normal variation over time. For example, body weight seems like a very reliable measure, but 5% variations are common even if the average weight has not changed at all. What this means is that any two measurements can vary by 5-10% in as little as a week without reflecting real change. This problem is particularly acute in chronic conditions where it is well recognized that patients have good and bad days. If you happen to measure a patient on a good day at the start and a bad day at the end, it may look like an effective treatment is not working. To overcome this challenge, clinical trials need to increase the sample size and lengthen the trial period. Both of these approaches dramatically slow the trial down and have major implications for the ROI.
This challenge becomes even worse when the measure has a subjective component. One of the most common outcome measures, patient reported outcomes (PROs), is subjective and suffers from many kinds of bias, including: collection mode, non-response, proxy/caregiver, recall, language, timing, and fatigue. Even apparently objective measures, such as the six-minute walk test, have issues with collection mode and tester bias.
This is not to say that we should not collect these measures. PROs, in particular, are important to reflect how the patients perceive the impact of the treatment. However, if they are used as primary endpoints, we condemn ourselves to larger and longer trials. What we need are continuous, objective measures that are important to patients.
Digital Health Measures
Digital health measures tend to be both objective and continuous. They measure quantifiable metrics such as acceleration, photoplethysmography (PPG) and temperature on a continuous or near continuous basis and provide insights into the full life experience of the patient. These metrics can be translated into assessments that are important to patients, such as activity levels and sleep quantity and quality. In other cases, they are translated into important health indicators, such as heart rate, heart rate variability, oxygen content, blood pressure, glucose levels, and many more.
The potential power of these digital health outcome measures can be seen in a case where they were applied as the primary endpoint in Bellerophon Therapeutics’ Phase 3 trial. Bellerophon used the change in daily activity measured by a wrist-worn acceleration sensor as the primary endpoint in its INOpulse trial for the treatment of fibrotic interstitial lung disease. With this measure, Bellerophon was able to reduce its sample size from 300 patients to 140 patients. Not only did this save the direct cost of enrolling the extra 160 patients, but it shortened the trial by approximately 22 months (based on the average enrollment rate for the trial).
The value of that reduction in time for a clinical trial is immense. According to Tufts CSDD, the mean daily value of prescription sales for a new drug is about $800,000, suggesting that a 22-month shortening of the trial could be worth somewhere in the range of half a billion dollars. Even more important, it could get valuable treatments to patients much sooner.
Barriers to Digital Endpoints
With such benefits, the question arises as why such examples are so few and far between. The answer comes down to three factors: time, money, and burden. Time is a major factor in that the digital endpoint needs to be validated prior to the start of the clinical trial for it to be considered a primary endpoint and generate the sample size reduction that generates these huge benefits. If the data does not already exist, the process of collecting that data must begin much earlier – generally years before the clinical trial starts – and often before the sponsor knows if they are even going to run a clinical trial.
Even though the potential financial benefits are huge, money becomes an issue when it needs to be spent years before a trial starts. In today’s environment, every organization has financial challenges. Spending resources for an uncertain value that is potentially years in the future is difficult.
The time and money challenges are real, but the greatest challenge may be burden – patient burden, CRO burden, and clinical site burden. Clinical trials are becoming more and more complex and adding one more element – even if it is relatively simple – may be perceived as an increased risk to the trial. Few sponsors are willing to do that.
As a result of these barriers, progress on digital endpoints has been extremely slow. The basic sensors (acceleration and PPG) on which many, if not most, of the digital endpoints are based, have been widely available in a wearable form for more than 30 years. And yet, Bellerophon has been the only sponsor to employ a digital endpoint as the primary endpoint in a pivotal Phase 3 trial in the U.S. And this lack of progress is not limited to clinical trials. Aside from continuous glucose monitoring, there has been almost no penetration of these advanced digital measures into the healthcare system. Something needs to change.
What can be Done Differently?
All the challenges described above can be solved if the right data are available. Doing the analytics is easy, quick, and cheap. Collecting data is what costs time and money and creates burden on participants, CROs, and clinical sites.
It may seem crazy to say that the problem is data when there are literally hundreds of millions of sensors (billions in the case of smartphones) generating data continuously. Unfortunately, almost all of that data is useless for validating measures in healthcare. It is based on proprietary measures which frequently change. Furthermore, it is in the hands of tech companies that guard their data and prevent it from being used by others. In addition, the population is not targeted and there is a dearth of validated outcomes, making the data far less useful.
All is not lost, however. There are many ways to improve what is being done today. Some examples include:
• Reuse the useful data that does exist – Just because most of the data generated is effectively useless for researchers, doesn’t mean it is all useless. Literally thousands of studies are done each year by academic researchers. Most of these studies receive grants which require that the data be made available to others. These data are high quality and generally connected to some outcome measures.
• Collaborate on data collection – Sponsors are unlikely to gain economic benefit from developing proprietary digital endpoints. In fact, collaborating and agreeing on standard endpoints is likely to make regulatory acceptance much easier. By collaborating, sponsors can share costs and management time, making it easier to justify research that is not directly tied to a trial.
• Collect more data on a routine basis – High quality medical research is being done on a continuous basis. Leveraging that research to do exploratory digital health studies alongside the main research provides an excellent opportunity for collecting targeted data with low risk. In particular, digital health technologies may be able to support label claims and be a useful tool for Phase 4 research.
• Reduce burden – Vastly expanding scientific data collection will not be possible without dramatically reducing the associated cost and burden. This is possible. Millions of people pay to wear smartwatches. Research sensors need to provide similar value to the wearer. Consumer devices are deployed without the customer needing to go to a clinical site. That can be done with research devices as well. There is no need to burden sites and CROs by adding digital devices that can be deployed and monitored remotely.
If the right data are collected, validation of digital endpoints and their implementation into healthcare becomes relatively straightforward.
A Final Word
This article extensively discusses the economic imperatives of adopting digital health technologies. That emphasis was chosen because the reality of the current system is that unless an approach works economically, it is likely to get little traction. The most important focus should be on the patient. Accelerating adoption of digital health technologies will speed the introduction of new medicines, improve treatment management, and eventually enable predictive interventions to prevent people from getting certain diseases in the first place. These are the true potential benefits to all who are potential patients (i.e., everyone).
References
1. https://www.norstella.com/why-clinical-development-success-rates-falling
2. Measuring the Return on Pharmaceutical Innovation, April 2024, Deloitte, US
3. Zini and Banfi, Int. J. Environ. Res. Public Health 2021, 18, 12445. https://doi.org/10.3390/ijerph182312445
Author Bio
Geoffrey Gill, MS, is CEO of Verisense Health, a digital health technology and data company, and Co-founder of the Open Wearables Initiative (OWEAR). Verisense Health is dedicated to producing clinical-grade digital health data and solving digital health data access and reuse problems. Geoffrey joined Verisense Health from Shimmer Research, the global wearable technology provider, where he served as President of Shimmer Americas. He received his MS in Management of Technology from the MIT Sloan School of Management.
