Across the past so many years, there have been pharmaceutical companies that have been integrating artificial intelligence steadily into many elements of clinical development. The effect that artificial intelligence has today is indeed being felt right from the bench to the clinic and even beyond.
As per the survey conducted by the Tufts Centre for the Study of Drug Development, there are one-third of the respondents that go on to support either full or partial execution of artificial intelligence in order to support clinical trial planning, its design and execution, as well as regulatory submission. According to the same survey, the AI usage results in average time savings of almost 18% when it comes to the implementation of clinical trials’ tasks and also activities.
It is well to be noted that since 2015, 75 AI-discovered molecules have gone on to enter the clinic, out of which 67 happened to be in ongoing trials as of 2023. Apparently, a watershed moment came into the picture in 2023 when a candidate from Insilico Medicine for the treatment of idiopathic pulmonary fibrosis – IPF – INS018_055 went on to become the very first drug discovered as well as designed by way of generative AI to enter the Phase II clinical trials gateway.
These instances just scratch the surface when it comes to the benefits of artificial intelligence to pharma. It was McKinsey that went on to identify 12 use cases that happen to illustrate the capacity of AI in order to largely enhance the quality, accuracy, and speed in the gamut of clinical trial development. The use cases, apparently, showed a lesser cost, fast enrolment, and a higher success rate due to the incorporation of AI.
Usage of Artificial Intelligence with Care
The ones who were surveyed by Tufts CSDD go on to state that having more successful executions and also use cases will enable driving the AI adoption. Nonetheless, this does not totally bypass the distinct barriers that have, as a matter of fact, hindered the AI’s adoption in the pharmaceutical landscape.
One of the most prominent challenges is the fear that artificial intelligence could very well go on to expose the sensitive data of the patient. It is worth noting that companies happen to be addressing this kind of risk in parts by way of applying anonymization as well as de-identification, data masking, and pseudonymization so as to eradicate the personal identifiable information from the datasets right before they are being used in various AI applications. Moreover, models that are trained on the basis of de-identified data get further safeguarded or encrypted in order to protect the patient-level data further. For example, the usage of more sensitive data points like date of birth can even be avoided by way of usage of proxies like age. One more issue happens to be the varied quality of the data that’s used so as to train the AI-led models. When it comes to clinical development, an AI model that is trained on bad data can elevate errors when it comes to forecasting as well as introduce more delays as far as trial timelines are concerned. Hence, it becomes imperative that companies make sure that data is indeed coming from sources that are trustworthy, have robust data management practices, and also undergo the quality assurance as well as the transformation that’s needed well in advance before being used for training the artificial intelligence models.
Human bias in the data takes place when an outcome or solution is either intentionally or unintentionally encouraged vis-à-vis the other one. If the AI model happens to be built or is trained on biased data, it can very well perpetuate these sorts of biases in its inferences and, in a way, can as well worsen the present inequality. In one of the studies, where the researchers went on to train purposefully an AI assistant with biased data, the precision of the diagnoses of the assistant went on to see a slump of 11.3%. IBM states that the awareness of bias has to be built within each data processing step and that the ongoing tracking as well as testing with the real-world data can even go on to catch as well as correct the bias much before it becomes well settled in the AI model. For instance, data points like ethnicity or even race can only be made use of as a data filter in order to identify participants who are eligible and if the protocol happened to be restrictive from an epidemiological point of view. As a matter of fact, such kinds of data points shouldn’t be used as predictors when it comes to the clinical trials operational metrics, which should be completely based on certain objective, measured, and also more epidemiological criteria that are relevant, like diagnoses, for instance.
It is well worth noting that even the pharma companies need to make sure that their compliance tools are on par so that they can make sure their AI applications stay within the boundaries of regulation. This can indeed be challenging, as some regulatory guidelines, like the Good Machine Learning Practice by the FDA, happen to be lagging the fast advancement of artificial intelligence, for example, taking into account the pervasive usage when it comes to generative AI. By way of making use of up-to-date as well as complete tools so as to assess, shape, and also track the data and also AI models, it will enable the AI with robust regulatory guardrails.
Smooth Sailing Ahead
The fact remains that pharma will continue to take AI in its stride in 2025, with organizations likely to embrace its usage at every phase of the process of development. As per the Fairfield Market Research forecast, the global market for artificial intelligence within the pharma spectrum is expected to reach a revenue of over $4.45 bn by the end of 2030 by way of having a robust CAGR of a touch above 19% from 2023 to 2030. This roll-out of AI is likely going to be encouraged by the new Trump administration, which has already made its stance clear on artificial intelligence. Shortly after taking office, US President Trump announced a $500 bn JV between Oracle, OpenAI, and SoftBank, which is going to invest in the infrastructure of AI. The CEO of Oracle, Larry Ellison, went on to suggest that a part of the project is going to be linked to the digital health records and also embraced the potential of AI for the development of new treatments for diseases as complex as cancer.
Some of the ways in which pharma firms are going to be ending up using AI tech in 2025 happen to include screening compounds as well as evaluating their epidemiological suitability and also their potential for repurposing, also planning along with optimizing clinical trial design and implementation, enhancing diversity and also inclusion in the enrolment of clinical trials, and at the same time bettering and streamlining the process of regulatory disclosure.
Other ways in which pharma companies will make use of AI in clinical trial planning as well as optimization include predictive along with prescriptive modelling, identifying the drug candidates, and also repositioning the approved drugs in order to treat other indications as well as designing and also optimizing the clinical trial protocols. Protocol design, especially, happens to be seen as a very promising frontier when it comes to using AI in the gamut of drug development, given the fact there is a time strain and also efforts needed so as to write trial protocols and also the delay brought by protocol alterations in terms of clinical trial timelines.
Companies such as Merck & Co. happen to be already making use of AI in the development workflows, such as assisting and also speeding up medical writing, and in a way intend to leverage the AI agents in order to automate the repetitive tasks like data clearing or even preliminary analysis. As sponsors go on to conduct clinical trial enrolment, the future is going to witness more pharma companies making utmost use of AI so as to enhance diversity and also inclusion, in particular in trials that happen to be diverse by way of design and also odds of finding patients who happen to fit an acceptable set of inclusions and also exclusions so as to speed up the trial timelines.
For instance, rather than waiting for the patients to get enrolled for the trials, Johnson & Johnson is making use of AI so as to locate clinical research sites and also investigators who are not just eligible but also suitable patients who could benefit from the J&J drug that’s being studied. Apparently, J&J also happens to be using data as well as AI so as to diversify clinical trials by checking out providers where the diverse patients are most likely to get treated and also giving priority for the enrolment of patients who are eligible from those providers. The usage of AI in regulatory applications is yet another area that is most likely to see an expansion in 2025. Pfizer is another pharmaceutical giant that plans to make use of ML-led analysis so as to identify which requisitions from the government regulators they may have and, in a way, prepare the answers to queries ahead of time, thereby saving weeks of lengthy discussions. The company is also looking into the use of AI when it comes to automating the production of massive amounts of reports and documentation that’s a prerequisite by the regulators.
