BOSTON (WHN) – Artificial intelligence is increasingly being positioned as a tool to personalize cancer treatment, analyzing vast datasets to predict disease risk and guide therapy. This technological shift, however, arrives after decades where physicians relied on established guidelines and, at times, intuition to navigate complex cases.
The personal experience of George Beauregard, DO, an internal medicine physician and author, offers a retrospective look at the challenges of individualized cancer care before the widespread integration of AI. In 2005, Dr. Beauregard, then 49, experienced painless hematuria, a symptom that, despite lacking common risk factors for kidney or bladder cancer, raised immediate concern. His medical history revealed a familial predisposition: both adoptive parents had developed urogenital cancers.
An ultrasound identified a lesion in his bladder. A subsequent cystoscopy confirmed a mass, described by a urologist colleague as “pretty angry-looking,” suggesting it was likely not benign. A transurethral resection of bladder tumor (TURBT) followed, with pathology revealing high-grade urothelial carcinoma that had invaded deeply into the bladder wall, showing multifocal lymphovascular invasion. At the time, the five-year survival rate for stage II muscle-invasive bladder cancer was approximately 45 percent.
Dr. Beauregard noted the anomaly of his diagnosis, as bladder cancer is overwhelmingly an age-related malignancy, with a median age of incidence in the septuagenarian years. The diagnosis prompted consultations with multiple specialists. Three urologists independently recommended a radical cystectomy with reconstruction. Medical oncologists, however, presented a divergence of opinion regarding chemotherapy regimens. One commented that the “wolf is already out of the cage,” indicating a high probability of microscopic disease beyond the bladder.
The available chemotherapy options, described by Dr. Beauregard as “one-size fits seventy-year-olds,” varied in their combinations of agents and timing relative to surgery. The lack of definitive evidence guiding the choice for his specific situation led to what he characterized as a “dartboard toss,” with his decision ultimately influenced by intuition. He did, however, opt to include trastuzamab (Herceptin) in his regimen after learning his cancer cells had a subclone of HER2-amplified cells, a decision driven by a potential survival advantage rather than academic curiosity.
This experience, he suggests, highlights the limitations of medical decision-making when faced with an “N-of-1” situation, where individual patient characteristics fall outside typical presentations or established treatment pathways. The mid-2000s saw hundreds of thousands of new research publications indexed on PubMed annually, requiring oncologists to synthesize information from clinical guidelines, randomized controlled trials (RCTs), meta-analyses, and trial registries.
The current landscape of cancer care, Dr. Beauregard observes, is evolving significantly. Today’s oncologists have access to a wider array of advanced diagnostic and therapeutic tools, including next-generation sequencing (NGS), circulating tumor DNA (ctDNA) and cell-free DNA (cfDNA) assays, CAR-T cell therapy, and various proteomic and transcriptomic profiling technologies. Yet, the challenge remains in determining precisely which of these sophisticated tools will best serve a patient’s unique cancer profile.
This is where artificial intelligence, particularly large language models, is presented as a potential solution. These AI systems are designed to process and synthesize immense volumes of diverse clinical data. The aim is to predict treatment outcomes, identify potential therapeutic dead ends, and tailor recommendations for individual patients with greater speed and accuracy. The concept is to move from broad, generalized approaches to more precise, iterative interventions, akin to using “fine scalpels, not blunt instruments.”
While Dr. Beauregard is a survivor of his 2005 diagnosis, he reflects on how a data-driven AI platform might have offered a different recommendation for his specific case. The potential for AI to analyze complex, individualized data points and inform treatment decisions holds promise for improving patient outcomes, though he acknowledges that perfection in any system is unlikely. An ongoing critical area of focus, he emphasizes, remains the advancement of early cancer detection, which could offer significant hope for better prognoses.