Clinical trials have been an integral part of medical progress for decades, serving as a crucial conduit for advancing medical treatment plans and expanding the knowledge surrounding such. The urgency to achieve successful outcomes within clinical trials, particularly in the realm of pharmaceutical development, continues to surge to unprecedented levels, and the dedication of physicians is no exception. Now, the integration of artificial intelligence (AI) is reshaping the imaginability of clinical trials. AI aims to enhance efficiency and precision while introducing a nuanced dimension of consideration to clinical practice, both in the patient and medical professionals’ experience. In fact, as of October 7, 2022, the FDA had approved 521 medical devices grounded in AI/machine learning. The influence of AI reverberates across clinical trials through its role in optimizing the management of patients, enhancing participant selection, and providing a deeper insight into medical conditions, particularly in oncology. 

Within the realm of cancer clinical trials, which often stand as one of the most prominent sources of advancements in the medical field, participation rates are meager and hold uncertainty in accuracy, leading to the cessation of clinical trials. When considering cancer clinical trials, less than 5% of patients choose to enroll in these trials. Furthermore, the statistic of one in every five clinical trials encountering termination due to inaccuracies sustains the need for strategic interventions to address these issues. 

AI has emerged as a tool to augment clinical trials and mitigate longstanding challenges– including inaccuracies. Typically, by leveraging large datasets, AI algorithms possess the ability to predict treatment responses and facilitate personalized patient care. In one AI-based clinical trial centered in Australia, it allowed for a 95.7% accuracy rate for exclusion and 91.6% accuracy for overall eligibility assessment. However, such success depends upon the accuracy, diversity, and quality of the data that it utilizes to train. Thereby, AI’s potential to inadvertently perpetuate inaccuracies and showcase biases that are easily overlooked is a concern that can not be ignored, especially in the context of cancer subtypes. 

Cancer encompasses various subtypes, each characterized by its own molecular, genetic, and clinical features—presenting a unique challenge. The subtypes commonly exhibit characteristics that are not present in more generalized cancers that would otherwise respond differently to treatment. The crux of the matter is made known when cancer subtypes suffer in representation in the training data utilized to develop AI algorithms for clinical trials; the resulting AI models are then at risk of not capturing the intricacies specific to the subtype(s) and in causation fail to deliver precise insights. 

Sarcoma, in particular, is a cancer that differs in crafting challenges to researchers and physicians due to its heterogeneity. The challenge held by physicians to identify therapies and allocate diagnosis to subtypes of sarcomas continues to heighten due to it holding over 100 varying subtypes. The rise of AI to sarcoma clinical trials will directly allow a new understanding of such subtypes while simultaneously allowing treatment opportunities to arise. The differentiating factor of the challenge sarcoma subtypes bring is not solely due to the quantity of subtypes but instead to what each subtype brings. Dr. William D. Tap, MD, of Memorial Sloan Kettering Cancer Center, New York, suggests that there is a “gathering of more than 100 to 150 different subtypes, each with unique biology.” Furthermore, with sarcoma often resting in soft tissue that gives a setting to a heterogeneous group of malignancies and the same malignancies also arising from bone tissues, it becomes more complex for physicians to provide a precise, accurate, and time-effective diagnosis and format a treatment plan. Directly, through its three main clinical tasks—detection, characterization, and monitoring of tumors—AI can assist in combating the challenge to improve the efficiency and accuracy in physicians’ decisions. 

Although AI has the ability to continue transforming the healthcare industry, there is a considerable concern that AI will replace the human touch, including the job of physicians. However, AI-based solutions will only optimize the human touch towards efficiency. Already, NCI’s The Cancer Genome Atlas Program, which encompasses over 3,600 people in their study, showcases that. There are additional reports on the rise of deep-learning models that are outperforming previous time-consuming methods which relied on expert annotations. These models now have the ability to predict disease outcomes through digitized whole-slide images, specifically of mesothelioma or hepatocellular carcinoma, and to predict metastatic relapse in patients with Soft Tissue Sarcoma.  

While AI will likely continue to transform clinical trials, one must recognize the bias that accompanies the acceleration. For patients of African and Asian ancestry, their genetic testing has shown to be present less in training models that are essential for predicting the effects of cancer treatments, as compared to patients of European ancestry. HSY also tested an AI model for breast cancer that gave rise to the same bias— predominantly white populations were more adequately represented in the training model compared to others, such as Hispanic women or those with a prior history of breast cancer. 

While the integration of AI has undoubtedly revolutionized the medical field, the necessity of physicians remains irreplaceable. Through AI’s capabilities, it is consistently maintaining a role in enhancing efficiency and accurate verification as it pertains to clinical trials. AI systems process vast amounts of medical data, diagnosing ailments and suggesting treatment options rapidly; however, they lack the intricate human touch that accompanies patient care. Physicians bring the unique values that AI lacks: empathy and the ability to consider patients holistically. However, AI’s role in optimizing the physician’s time and highlighting areas the human eye often overlooks can also not be ignored. Furthermore, physicians are essential for interpreting AI-generated solutions and ensuring that a personalized approach continues amid AI bias. The synergy held between the compassion of skilled physicians and AI’s capabilities will continue to define the future of healthcare, guaranteeing optimal patient care that doesn’t lose touch.