Physician-Complementing Artificial Intelligence in Oncology

Junren Chen, MD, PhD, MBA

Robert Peter Gale, MD, PhD, DSc (hc), FACP, FRCP, FRCPI (hon), FRSM

Use of artificial intelligence (AI) in oncology is advancing rapidly. AI was first used for reading radiology images and analyzing pathology slides. More recently, use of AI has expanded to analyzing large clinical data sets (big data). The next envisioned role for AI in oncology encompasses many spheres of physician activities such as choosing the best therapy and/or prescribing drugs.

Physician-Substituting vs Physician-Complementing AI

In these new spheres, AI can be physician-substituting or physician-complementing. Physician-substituting AI replaces all or portions of human faculties. Sometimes this is useful. For example, since the SARS-CoV-2 pandemic, for simple medical conditions, many Chinese are receptive to visiting an e-pharmacy platform that connects them to a physician who writes a prescription after a brief online chat and then delivers medicines to their homes’ front doors; that is, they do not seek help from a doctor they normally visit.¹ However, physician-substituting AI risks replacing highly skilled cancer physicians with those less skilled or even unskilled with the plausible net result of lowering the average human intelligence at all levels of and decreasing the quality of cancer care.²

This raises the question: What is the goal of AI in oncology? Is it to raise the performance of less skilled or unskilled oncologists to near that of skilled oncologists, or something else? Different AI experts will have different answers, but it is wise to remember the idiom: A rising tide lifts all boats.

Physician-complementing AI empowers skilled physicians to do more and makes them even better at what they do. An example is ambient documentation technology.^3,4 Regardless of their degree of empathy skills, physicians spend considerable time during a patient encounter with nonpatient-facing tasks like documenting the visit in the electronic medical record (EMR) and (re)ordering laboratory tests and drugs. AI scribes can assist these tasks facilitating empathetic physicians to engage in greater doctor-patient interactions.

Assistance With Clinical Decision-Making

Another area where even highly skilled oncologists may need help is navigating the enormous complexity of deciding benefit-to-risk ratio of a potential therapy for a specific patient. In addition to expert consensus statements, clinical practice guidelines, and results of randomized controlled trials, experienced oncologists consider many other variables including their experience with similar patients and therapies (recall heuristic), concern for quality of life, doctor-patient interactions, and confidence in their ability to decide on a therapy plan.⁵ As more and more variables are considered, the curse of dimensionality quickly arises such that only exceptionally skilled physicians, if any, can integrate all of these variables. Physician-complementing AI has the potential to help physicians improve on these deliberative processes.

Physician-complementing AI empowers skilled physicians to do more and makes them even better at what they do.

— JUNREN CHEN, MD, PhD, AND ROBERT PETER GALE, MD, PhD

Tweet this quote

Of note, physician-complementing AI can help oncologists evaluate many clinical covariates. Most people cannot consciously, simultaneously process many variables because conscious thoughts have a limited bandwidth.^6,7 Because of the low capacity for conscious deliberation, decision-makers often focus on a subset of variables they believe most important. Weighting of variables is often imperfect, distorted by anchoring and recall bias heuristics.

For example, most skilled physicians cannot process all available clinical and laboratory data when deciding whether to prescribe a drug unless an AI agent is embedded in the EMR to help them.⁸ Without AI assistance, they often rely on only a few variables and may make a suboptimal decision or defer a decision. In our RAND-Delphi study of breast cancer therapy, physicians said they needed to know about 15 covariates before recommending a therapy. When we analyzed the data using recursive partitioning, we found they used 3 or 4 at most.⁹

A smart, AI-powered interactive agent coupled with analyses of previous medical records could identify a person’s weighting of outcomes measures.

— JUNREN CHEN, MD, PhD, AND ROBERT PETER GALE, MD, PhD

Tweet this quote

With so-called precision oncology, the number of variables to be considered in therapy decision makes this even more challenging. Ideally, each person’s therapy should consider their unique features such as genetic variant topography, actionable variants, comorbidities, social milieu, and therapy goal(s). Determining how to accomplish this is complex and challenging.

For example, few people resemble subjects enrolled in the clinical trials on which expert consensus statements and clinical practice guidelines are based. When prescribing an immune therapy drug for someone with cancer who has an autoimmune disorder, it is important to balance benefits and risks. There are few data to inform decision-making in this setting.^10,11 Physician-complementing AI can help oncologists by summarizing evidence from diverse data sources and EMRs of similar people receiving immune therapy, thereby providing structure in considering huge numbers of variables informing decision-making.

Physician-complementing AI can also help oncologists and patients prioritize therapy goals. Efficacy endpoints in oncology trials are typically progression-free survival and/or overall survival. Expert consensus statements and clinical practice guidelines assume these endpoints are what people value most. However, many people may rank other outcomes measures like quality of life above progression-free or overall survival. Others may prefer a therapy associated with less time spent in health-care facilities compared with another therapy associated with longer survival but more time spent in health-care facilities.¹² Yet others may prioritize avoiding pain and/or avoiding adverse events over longer progression-free or overall survival.

Although physicians can theoretically adjust therapy decisions based on these considerations, in practice this is difficult to do.¹³ AI has the potential to assist physicians in this task. A smart, AI-powered interactive agent coupled with analyses of previous medical records could identify a person’s weighting of outcomes measures, and another AI agent could rank available therapy options accordingly. The recently announced collaboration of ASCO and Google Cloud to develop an AI-based tool for ASCO guideline queries is a promising step in this direction.¹⁴

Transforming Cancer Care

There is consensus that AI has the potential to transform cancer care. We look forward to hearing about the new ideas discussed at the 24th Cancer Centers Workshop, “Precision Oncology Powered by Data Intelligence.” With modern cooking techniques (eg, sous vide, flash freezing, and ultrasonic homogenizing), master chefs create new, previously unthinkable dishes. We suggest physician-complementing AI will allow oncologists to take cancer care to new heights.

DISCLOSURE: Dr. Chen reported no conflicts of interest. Dr. Gale is a consultant to Antengene Biotech LLC and Shenzhen TargetRx; Medical Director of FFF Enterprises Inc.; has served as a speaker for Janssen Pharmaceuticals, BeOne Medicines (formerly BeiGene), and Jiangsu Hengrui Pharmaceuticals; has served on the Board of Directors of the Russian Imperial Foundation for Cancer Research and StemRad; and acknowledges support from the UK National Institute for Health and Care Research.

REFERENCES

1. Lin Y, Xu X, Liu Y, et al: Perception and acceptance of telemedicine use in health care among the general public in China: Web-based cross-sectional survey. J Med Internet Res 26:e53497, 2024.
2. Feinstein AR, Sosin DM, Wells CK: The Will Rogers phenomenon: Stage migration and new diagnostic techniques as a source of misleading statistics for survival in cancer. N Engl J Med 312:1604-1608, 1985.
3. Tierney AA, Gayre G, Hoberman B, et al: Ambient artificial intelligence scribes to alleviate the burden of clinical documentation. NEJM Catal Innov Care Deliv 5(3), 2024.
4. Olson KD, Meeker D, Troup M, et al: Use of ambient AI scribes to reduce administrative burden and professional burnout. JAMA Netw Open 8:e2534976, 2025.
5. Nilsson MS, Pilhammar E: Professional approaches in clinical judgements among senior and junior doctors: Implications for medical education. BMC Med Educ 9:25, 2009.
6. Halford GS, Baker R, McCredden JE, et al: How many variables can humans process? Psychol Sci 16:70-76, 2005.
7. Dijksterhuis A, Bos MW, Nordgren LF, et al: On making the right choice: The deliberation-without-attention effect. Science 311:1005-1007, 2006.
8. Chen J, Cao Y, Feng Y, et al: Autonomous artificial intelligence prescribing a drug to prevent severe acute graft-versus-host disease in HLA-haploidentical transplants. Nat Commun 16:8391, 2025.
9. Gale RP, Park RE, Dubois R, et al: Delphi-panel analysis of appropriateness of high-dose chemotherapy and blood cell or bone marrow autotransplants in women with breast cancer. Clin Transplant 14:32-41, 2000.
10. Yue Q, Deng X, Xu Y: Limitations in quantitative harm-benefit assessment in immuno-oncology. JAMA Oncol. October 9, 2025 (early release online).
11. Heyward J, Segal JB: Limitations in quantitative harm-benefit assessment in immuno-oncology-reply. JAMA Oncol. October 9, 2025 (early release online).
12. Gupta A, Eisenhauer EA, Booth CM: The time toxicity of cancer treatment. J Clin Oncol 40:1611-1615, 2022.
13. Wiercioch W, Morgano GP, Piggott T, et al: GRADE guidance: Using thresholds for judgments on health benefits and harms in decision making (GRADE Guidance 42). Ann Intern Med. October 7, 2025 (early release online).
14. Hudis CA, Kurian T: ASCO and Google Cloud set forth a vision for using AI to modernize health care and advance oncology. The Cancer Letter 51:10-11, 2025.

Dr. Chen is Professor at the Institute of Hematology, Chinese Academy of Medical Sciences, Tianjin. Dr. Gale is Visiting Professor of Hematology at the Imperial College London.