The process of bringing a drug to market is not easy, and that’s an understatement. The average clinical trial takes between 10 to 15 years to complete, costs about $2 billion, and has a success rate of about 10%. However, clinical trial high failure rates persist despite these substantial investments.
Additionally, 86% of all research clinical trials fail to meet enrollment deadlines due to difficulties in selecting and recruiting patients to participate in clinical trials. Because of these challenges, companies are constantly looking for ways to make the process easier, cheaper, and safer. Using new technologies can help streamline clinical trials in many ways.
Over the past few years, artificial intelligence (AI) has become an increasingly important part of the pharmaceutical industry. Artificial intelligence refers to the development of smart machines that can perform tasks usually performed by humans. As in other sectors, AI technology has the potential to revolutionize clinical trials for drug development as well.
There are several ways AI can help organizations, such as selecting patients by reducing heterogeneity, taking cohorts, assisting with recruitment, and monitoring patients during clinical trials. It can also help shorten trial cycles, keep track of potential patient dropouts, and improve patient outcomes.
However, AI’s full potential may not be realized for a decade or more, but machine learning technology is presently technology. Machine learning (MI), a subfield of artificial intelligence, involves using computer algorithms to interpret raw data and knowledge derived from that data for various purposes without human involvement.
Moreover, clinical trials also encounter big data challenges. A human can create rules for classifying a small dataset and achieve accurate results. It is theoretically possible to handle even a few thousand records or tables. But what happens when you add multiple variables and millions of data to the mix? It becomes impossible to perform everything manually and achieve accurate results. In recent years, the number of data points collected in a clinical trial has increased, and one study notes a 40% growth in data collected yearly.
Thus, AI and machine learning are ideal solutions for the industry’s big data problem. By integrating, mapping, and transforming multiple datasets into one common data model, machine learning can help enterprises control scalable data management. The machine learning approach unifies data as it comes in. With algorithms that match and link incoming data with other datasets, all business units have greater access to enterprise data. The result is a faster, more reliable, and more scalable analytics system.
Elvin Thaulund, clinical business analyst consultant and the director of industry strategy at Oracle Health Sciences, perceives AI and machine learning as new technology. This telescope will enable us to see into the future. This will be possible due to the predictive capabilities of this technology.
Many obstacles can cause trial failures, increasing the cost and time of bringing a new treatment to patients. With machine learning, scientists can address those obstacles and predict enrollment before they begin writing their first protocol using a traditional trial design. Frequently, they write a protocol only to find out afterwards that they cannot enroll the patients they anticipated. Through machine learning, they can predict enrollment ahead of time and then design the proper protocol based on the results.
We already mentioned that the traditional approach to clinical trials and drug development fails for numerous reasons, including under-enrollment, inconsistent data, mid-trial attrition, and unexpected side effects. Traditional clinical trials cannot develop complex new therapies because they lack analytical sophistication, flexibility, and speed.
Patient recruitment is one of the major setbacks to clinical trials. This is especially true for those targeting smaller, heterogeneous patient populations. Each trial starts with a clinical plan and the development of a protocol where each site needs to supply the number of potential patients they can recruit.
Some manage to recruit the number they promise, but approximately 50% struggle with under-enrollment and 10% don’t manage to recruit a single patient. Unfortunately, many trials are cancelled precisely due to the insufficient number of recruited patients.
On top of that, trials are time-consuming and expensive. Aside from the $2.6 billion needed to bring a new treatment to patients, approved drugs must also cover their own expenses and costs associated with failed treatments.
Thanks to machine learning technologies, it will be easier to predict the outcomes of clinical trials, thereby reducing drug approval times, lowering costs, and helping to develop new treatments more quickly. In addition to reducing uncertainty, more precise predictions can increase the amount of money that investors are willing to pay for clinical trials. Plus, this will facilitate patient enrollment, recruitment, and retention to a much greater extent.
When developing a new treatment, machine learning could help researchers accomplish the following:
If a pharmaceutical company implements all the factors we discussed previously, trial results will turn out better, and trials will be less consuming with fewer failures. Furthermore, trials would be less expensive, as would the associated prices of medicines. Also, this will reflect the cost of launching a new drug to the market by lowering the $2.6 billion average cost by approximately one-third.
However, the existing business structures and processes are currently preventing the realization of many of these potential savings. Instead of applying machine learning before the clinical trial onset, it is usually utilized too late, once critical decisions are made. These actions lead to rescue studies that are easily prevented since machine learning is already available.
We have already explained that each trial starts with the clinical development plan. It is precisely this stage, the portfolio level, where machine learning needs to be applied. Only then will the researchers be able to develop more detailed design decisions. Study designs that are more likely to fail would be rejected or redesigned using machine learning in the planning process. For instance, a decentralized study design may work better for inappropriate trials for traditional study designs.
But since machine learning is already a reality and stakeholders are ready to invest in trials, you might wonder where the problem lies. Some key opinion leaders say it is in human understanding. Since researchers and clinical planners are humans, they are the first who need to understand the importance of AI and machine learning, ask the right questions, and understand what data is required.
Overall, AI and machine learning have already become indispensable assets to healthcare and life science technology companies. In addition to guiding trial development, improving data, and ensuring accuracy, they provide transparency. Scientists can increase trial success rates and help other researchers worldwide by incorporating these resources into research processes. When clinical trial data is easily shared and analyzed, it benefits everyone. Although there are challenges to overcome, there is no doubt that machine learning and AI will improve clinical trials in the future.