Cancer Care by Committee to be Superseded by Personal Physician–Patient Partnership Informed by Artificial Intelligence

Multidisciplinary Tumor Boards

“Multidisciplinary teams are an alliance of all medical and health care professionals related to a specific tumour disease whose approach to cancer care is guided by their willingness to agree on evidence-based clinical decisions and to co-ordinate the delivery of care at all stages of the process.” ²

“It appears intuitive that MTBs are essential to clinical decision-making and patient management; however, it is unclear whether that belief is supported by evidence that they improve patient outcomes.” ³ Systematic review of multidisciplinary tumor board (MTB) shows evidence that “they lead to significant changes in the way cancer patients are assessed and managed. Some of these alterations are deemed to be beneficial as they are in accordance with recommended clinical guidelines. However, it is unclear if various changes in assessment and treatment lead to clinically significant differences in patient experience or quality of life and there is only weak evidence of improved survival outcomes.” ⁴ Improved progression-free survival (PFS) and overall survival (OS) attributed to MTB advice are challenging to verify, since referred patients are often more complex and difficult to treat than nonreferred patients, and randomized controlled trials (RCTs) are not ethically justifiable. ⁵

Efficacy of MTBs in cancer care is claimed using a metric of change to management plans in the majority of referred patients, and the extent of adherence to guidelines and MTB recommendations. ^6,7 However, such MTB evaluation studies often do not follow-up the patients referred to them, and clinical outcomes in the form of OS are not available. Where long-term follow-up has been performed, no significant change in mortality was demonstrated over a 10-year period of observation of lung cancer patients referred to MTB versus those receiving a standard model of care. ⁸ However, after adjustment for possible confounders, mortality rates were reported to be lower in the MTB group. It was remarked that the noted improvements in outcome of the multidisciplinary model could not be reliably differentiated between a multidisciplinary clinic, tumor board conference, shared decision treatment plans, and monitoring and surveillance programs. ⁹

A definitive American Society of Clinical Oncology (ASCO) evaluation and review of MTB function over 25 years concluded that: “The evidence from studies on the impact is patchy, but, overall it supports their value; however, a continuous and rigorous approach to ensure their efficiency is essential.” ¹⁰

Recent Umbrella review of the impact of MTBs on cancer care concluded that evidence was still lacking with respect to survival rates, quality of life (QOL), and patient satisfaction. It was remarked that there was lack of patient involvement in the decision-making process. ¹¹ Indeed, patients are required to give informed consent for their cancer treatment to be decided by a faceless committee, whereas many would be more confident in reposing their trust in an individual, familiar, and caring oncologist who understands their personal anxieties, values, and needs.

An observational study of MTB decision-making in breast cancer concluded that exchange of multidisciplinary knowledge, deemed to be an essential aspect of meetings, does not always seem to occur in practice. “The extent to which recommendations are derived from consensus during MTBs remains unclear.” ¹²

Although the introduction of MTBs has increased the opportunity for professionals to be included in cancer treatment decisions, opportunity for the patient to be involved has been diminished. MTB decision-making may thus present significant barriers to engaging with patient values and preferences and, therefore, presents obstacles to delivering shared decision-making. ¹³

“No decision about me without me” expressed by patients may be contrasted with the remarks of an MTB doctor whose principal priority in discussion is disease-focused rather than person-focused: “I don't personally think the [MTB] is the place to bring up patient-centred data because, a number of reasons, the religious and cultural beliefs of the person without the person themselves are most likely to be misrepresented and pretty much meaningless without the human being there … you're not managing the patient in an [MTB], you are just making a decision on part of the information that needs to be analysed as a group, because ideally you want different people to look at the same information and come to a collective decision, but you don't need them to come to a collective decision on how this impacts on cultural beliefs.” ¹⁴

A survey of over 2000 cancer health professionals in the United Kingdom found that the majority felt it was neither desirable nor practical to include patients in MTB meetings. ¹⁴ Review of 16 MTB meetings disclosed that data concerning comorbidities and psychosocial context were almost always missing or superficially presented. Factors such as patient preference, social and financial status, and presence/lack of adequate caregiver were rarely discussed in the meetings and this was attributed to the lack of studies inquiring about the real weight of these factors in influencing treatment decisions. Almost all published articles agree on the fact that adherence to guidelines is one of the main factors that is encouraged in MTB discussions, but sometimes adherence to guidelines did not determine any change whatsoever in survival outcomes between patients discussed in MTBs versus those who were managed outside the setting of MTBs. ¹⁵

It has been said that “MTB members tend to overrate their performance in relation to patient-centered information, and it is likely that members are not aware of the extent that issues, such as comorbidity, are ignored.” Even when treatment decisions are reached in MTBs, the actual treatment received is more likely to be discordant with that decision among patients with comorbidities. ¹⁶

The relative absence of in-depth patient-centered discussion in real-world MTBs is understandable given the exponentially increasing patient load caused by the bureaucratic mandate that all cancer patients in many European countries, as in Australia, must be referred to an MTB for consensus management. The typical allocation for each patient to be discussed in MTB is little more than 5 min during which time the imaging, pathology, omics, and other technical data have to be presented to the 15–25 specialist attendees. ⁵ In addition to brief presentations, the radiologists and pathologists must collect, collate, and analyze all extant imaging and pathological/immunohistochemical and genomic data before the meeting. These data preparation and actual attendance at MTBs are very often a hidden cost in terms of time spent and lack of reimbursement, or even recognition, by hospital administration. In fact, up to 25% of members of MTBs may suffer psychiatric morbidity with high levels of emotional exhaustion in team leaders and nurses and feelings of low levels of personal accomplishment being prevalent in histopathologists and radiologists. ¹⁷

Most oncology MTB meetings could be considered legally as comprising formal patient referrals that give rise to a duty of care for each individual doctor. ¹⁸ Medicolegal recommendations in reference to each case discussed include identification of each member of the MTB meeting who contributed to the treatment plan and decision-making process, and their names recorded, as they each have a duty of care to the patient.

A clear distinction should be recognized between MTBs convened for consideration of molecular/chemotherapeutic/immunotherapeutic targeting of tumors and dedicated multidisciplinary clinical surgical planning meetings, such as those formulating specific operative treatment of head and neck cancer, where the ear nose and throat surgeons, plastic surgeons, radiotherapists, and medical oncologist are all active players in the procedure and follow-up therapy of a particular patient, and each specialist is personally responsible for a particular aspect of treatment.

At the other end of the spectrum, MTBs devoted to rare diseases, usually held in specialist regional or national academic referral centers, will continue to fulfill a need beyond the capability of local MTBs and remain outside the scope of the possible transition of hospital MTBs in response to exponentially increasing patient load and the prospect of an overwhelming tsunami of data.

Decision-making in oncology involves the consideration of a complex set of diagnostic, therapeutic, and prognostic uncertainties, potentially leading to disagreement about the best course of action. Most recommendations in oncology are not based solely on high-level evidence, and while guideline adherence in MTBs is lauded, only 6% of guidelines of the National Comprehensive Cancer Network (NCCN) were based upon high-level evidence. ¹⁹ An approach purely based on clinical algorithms and traditional evidence-based medicine (EBM) alone may be insufficient, especially when the limited availability of information and complexity of the patient and the environment are not sufficiently considered. ²⁰

In addition to overseeing conformity with EBM clinical trials, somewhat paradoxically, MTBs may recommend enrollment of patients on early phase Pharma RCTs, ¹⁰ all of which share the prime objective of drug approval and registration, and which are not designed to provide individual benefit to the research subject. ²¹ Such clinical trials, or expanded access programs and off-label indications, have limited evidence of efficacy in daily practice but may be seen by MTBs to offer valid treatment options when no others are available. ²²

Phase 1 clinical trial design is focused upon dose and schedule determination, patient safety, and limited patient exposure to ineffective doses of investigational agent. ²³ The position of the ASCO is that phase 1 trial enrollment potentially provides the “prospect of direct medical benefit,” and the U.S. Food and Drug Administration (U.S. FDA) policy acknowledges that the primary aim of phase 1 trials is “to gain early evidence of effectiveness.” ²⁴ However, the recent acceptance of Pharma risk-targeting design of Phase 1 clinical trials specifically seeks to find and focus upon a dose that is expected to cause rapidly emerging severe or life-threatening adverse events in a predetermined percentage of patients, usually 25%, to define dose-limiting toxicity. ²⁵ This is greater than the 16% who may be expected to experience any degree of clinical benefit. ¹⁶ Oncology drug development is associated with failure rates of around 90%, ²⁶ and the question “is participation in cancer phase 1 trials really therapeutic” ²⁷ is pertinent for MTBs when they recommend that patients enroll in Pharma clinical trials.

Multidisciplinary Tumor Board Guidelines and Theranostics

A comprehensive authoritative evaluation of radiopharmaceutical therapy (RPT) of cancer, published in Nature Reviews in 2020, ²⁸ remarked that “most pharmaceutical companies are not familiar with radiation and radionuclide aspects of RPT and the deployment of RPT agents for cancer therapy is also unfamiliar to the oncology community … RPT has been a treatment modality of last resort and available only in small clinical trials or as part of compassionate care from a small number of institutions in Europe [and Australia] and even fewer in the United States and the rest of the world. In the sense that RPT has no well-defined community of stakeholders it has been an ‘orphan treatment’ modality for many years … Compared with almost all other systemic cancer treatment options RPT has shown efficacy with minimal toxicity … Unlike biologic therapy it is far less dependent on an understanding of signalling pathways and on identifying agents that interrupt the putative cancer phenotype driving pathway(s). In contrast to biologics or chemotherapeutics both radiation delivery and the biologic response to radiation may be mathematically modelled and used to understand the parameters of a treatment that are most important in influencing efficacy and toxicity.” ²⁸

Apart from a few theranostic referral centers, nuclear physician expertise in cancer therapy is seldom represented on MTBs, and it is radiologists who provide image interpretation and analysis. The nuclear physician workforce is currently not sufficient to meet the growing demand for greater engagement with patient management, which requires a deep understanding of disease processes, pathology, pharmacology, and treatment algorithms, to enable application of radiotheranostics in the overall context of the patient's disease management. ²⁹

Despite strong recommendations to embed the nuclear physician into the MTB, ³⁰ they are under-represented in cancer consensus meetings. For example, the panel of the Advanced Prostate Cancer Consensus Conference 2021 excluded nuclear medicine experts from voting, on the grounds that they were not directly involved in clinical decision-making. ³¹ One of those nuclear physicians in attendance had been the Chief Investigator on the pivotal Phase 2 TheraP RCT of ¹⁷⁷Lu-prostate-specific membrane antigen (PSMA) versus cabazitaxel chemotherapy in metastatic castrate-resistant prostate cancer (mCRPC) ³² and evidently was accorded no vote in the specific discussion dedicated to ¹⁷⁷Lu-PSMA treatment and PSMA targeted agents in diagnostics and therapy. ³¹

The disconnect between nuclear physicians and medical/surgical oncologists is epitomized by the comparison of contemporaneous issues of Journal of Nuclear Medicine and New England Journal of Medicine in February 2018 specifically addressing prostate cancer. The Journal of Nuclear Medicine proclaimed “Why targeting PSMA is a game changer in the management of prostate cancer,” ³³ while the comprehensive review in the New England Journal of Medicine entitled “Metastatic prostate cancer” ³⁴ made no mention of clinical studies of ⁶⁸ Ga/¹⁷⁷Lu/²²⁵Ac-PSMA theranostics. A more recent authoritative review of precision medicine in prostate cancer in late 2020 in Nature Cancer also made no reference to ¹⁷⁷Lu/²²⁵Ac-PSMA theranostics. ³⁵

A 2022 review of real-world treatment of mCRPC across Europe and in Japan showed high compliance with national guidelines, which do not mention PSMA radioligand theranostics. ³⁶ Swedish guidelines of 2022 “strongly recommend that patients entering the castrate-resistant phase [of prostate cancer] are discussed in a MTB, if possible the patient should be offered to participate in a clinical trial.” ³⁷ Not only do these Swedish guidelines (which are said to mirror general European guidelines) omit consideration of ¹⁷⁷Lu/²²⁵Ac-PSMA therapy, there is also a notable absence of a general recommendation to use PSMA-PET/CT for staging of prostate cancer.

In March 2022 the U.S. FDA approved ¹⁷⁷Lu-PSMA-617 for PSMA-positive mCRPC as determined by ⁶⁸ Ga-PSMA-11 PET imaging, based upon the VISION Study. ³⁸ Long before this, specialized theranostic centers in Germany, Italy, and Australia, free from the constraints of arbitrary institutional guidelines, have been practising real-world ¹⁷⁷Lu-PSMA therapy of mCRPC over the past 12 years with long-term follow-up demonstrating meaningful clinical benefit of prolonged survival and enhanced QOL. ^21,39

Currently in the United States, an MTB decides that there will be no exceptions to be made for patients not having received both prior chemotherapy and novel androgen receptor axis inhibitor treatment, and eligibility decisions are made on PSMA PET/CT imaging studies performed up to 4 months before MTB consideration. ⁴⁰ Median duration between MTB review and administration of the first cycle of ¹⁷⁷LuPSMA was 59 days (32–129 days) in a real-world academic referral center practice, which accords no individualized treatment discretion to the therapeutic nuclear physician.

In respect of radioligand somatostatin receptor theranostics of neuroendocrine tumors (NETs) the delay between clinical trial demonstration of efficacy of ¹⁷⁷Lu-peptide receptor radionuclide therapy (PRRT) ^41,42 and U.S. FDA and European regulatory agency approval was almost two decades. Regulatory approval on the basis of the NETTER Pharma RCT ⁴³ “was like listening to the result of a football game being announced a week after we had watched the match” according to one of those of us in Italy, Holland, Germany, and Australia who had already been using ¹⁷⁷Lu-PRRT routinely for 20 years, ⁴⁴ and this sentiment was echoed by the founding father of clinical PRRT. ⁴⁵

In fact, the NETTER RCT Study potentially underestimated the efficacy of ¹⁷⁷Lu-octreotate PRRT. ⁴⁶ The reported clinical outcomes had already been superseded by the advent of combination radiopeptide therapy with radiosensitizing chemotherapy ^47,48 and by the introduction of ²²⁵Ac-octreotate alpha PRRT ⁴⁹ which have increased efficiency in recurrent and refractory NETs. An American patient who flew to Germany for ⁶⁸ Ga-DOTATOC PET/CT in 2008 and the next year went again for ¹⁷⁷Lu-DOTATATE PRRT lamented that “it took 22 years to bring forward a new approval for an imaging agent” and 18 years to approve an efficient therapeutic agent. ⁵⁰

While joint International Atomic Energy Agency, European Association of Nuclear Medicine (EANM), and Society of Nuclear Medicine and Molecular Imaging guidelines for PRRT of NET, published in 2013, ⁵¹ declared PRRT to be indicated for the treatment of patients with positive expression of somatostatin receptor 2, or metastatic or inoperable NET, non-nuclear medical oncologists were more dependent upon more conservative guidelines formulated by the European Neuroendocrine Society (ENETS) which in the absence of RCT evidence-base recommended PRRT only as a third-line, last resort, treatment. ⁵² Such guidelines, by the express admission of their authors, have effects upon MTB management which simply cannot be quantified, and they remark that “sometimes evidence is not necessary to notice the benefit.” An oncologist colleague paraphrased Sir William Osler to express his opinion that “He who treats cancer without an MTB sails an uncharted sea, but he who relies entirely upon an MTB management decision does not go to sea at all.” Nevertheless, the reported conformity with MTB recommended practice in the management of NETs according to ENETS Guidelines is >88%. ⁵³

To integrate PRRT therapies into routine appropriate clinical practice the EANM convened a multispecialty forum in January 2020 in close collaboration with ENETS to achieve consensus. ⁵⁴ Rapprochement was reached in terms of earlier treatment of receptor positive NETs as second line therapy.

Radioimmunotherapy (RIT) of Non-Hodgkin lymphoma (NHL) with ¹³¹I-tositumomab was approved by U.S. FDA in 2003, but despite demonstration of significant clinical benefit in RCTs it was not adopted by oncologists, and contemporary North American guidelines did not mention RIT or recommend its use, ⁵⁵ and ¹³¹I-tositumomab was withdrawn in 2014. State-of-the-art review in 2022 of novel treatments in B cell NHL fails to cite any RIT studies. ⁵⁶ Recent report of the phase 2 INITIAL Study of ¹³¹I-rituximab RIT in first line treatment of advanced Follicular NHL with over 10-year follow-up achieved 80% OS, ⁵⁷ and clinical trial of ¹⁷⁷Lu-Lilotomab Satetraxetan RIT of relapsed NHL ⁵⁸ was totally ignored by hematologic oncologists in their “comprehensive” review. ⁵⁶ RIT is thus unlikely to be incorporated into any official guideline for treatment of NHL and cannot be legitimately recommended by any MTB.

RIT given intraventricularly for advanced refractory central nervous system malignancies using ¹³¹I-omburtamab, targeting B7-H3, ⁵⁹ was considered under the FDA Oncology Real-World Evidence Program to inform regulatory decision-making, but approval was withheld on the grounds that a real-world external control arm was inadmissible for comparison of OS outcomes. ⁶⁰

(Virtual) Molecular Tumor Boards and Artificial Intelligence

Convening a regular meeting of busy medical oncologists and domain experts in a conference room to integrate different data types about a patient, including past medical and treatment histories, immune-pathology and imaging reports, and genomic, proteomic, radiomic, and molecular profiling results, and to make a management decision within a very limited time frame, is fast becoming unfeasible. The medical oncology system operates in an overburdened environment now, and the ability of oncologists to keep up with the ever-expanding lists of biomarkers, treatment-matching rules, cancer biology, and single and combination molecular and immunologic target therapies will become untenable in the face of the complexity of the input data and exponential increase in treatment options to be evaluated in respect of each patient.

The advent of next-generation sequencing and transcriptomic sequencing has enabled the rapid identification of potentially pathogenic molecular abnormalities, and many new agents with striking ability to modify the action of specific aberrant proteins have been developed. However, both a deep understanding of the mechanism of action of a drug and exploitation of molecular biomarkers, to choose the appropriate patients for treatment, are essential.

Giving targeted drugs to unselected patients is associated with paltry objective response rates of <5%. ⁶¹ Most immunotherapies have obtained regulatory approval for all comers, regardless of biomarker profile, which is antithetical to the key tenet of precision oncology, and which raises concerns over cost-effectiveness, modest magnitude of clinical benefit, and the benefit-harm ratio. ⁶²

While a virtual MTB platform might address logistic attendance issues ⁶³ the process remains complex, time-consuming, and labor intensive. The expectations and burden on MTB members are predicted to increase in the near future, given that it is physically impossible for a clinician to comprehensively examine the burgeoning specialty literature. Introduction of artificial intelligence (AI) to the MTB is now advocated to facilitate precision medicine using large-scale cancer clinical and biological information. ⁶⁴ However, human oversight will be critical to the adoption of novel data-driven technologies such as AI machine learning in cancer care. ⁶⁵

An MTB is the forum ideally suited to provide expert human oversight of AI applications in cancer management. Collegiate specialist-informed analysis of AI data has the potential to extract clinically relevant knowledge for appropriate real-world enhancement of patient care. This humanizing of machine-learning AI will inevitably become a demanding and time-consuming exercise and could become the new prime function of an MTB which may compromise the current core objective of prospective decision-making in respect of detailed management of each cancer patient.

The field of molecular oncology is rapidly evolving, and hence, the understanding and interpretation of biomarkers and outcomes are likely to change. One MTB reported successful facilitation of the interpretation of multiple testing modalities, including next generation sequencing, cfDNA, mRNA, and immune-histochemistry, in a precision N-of-One strategy and achieved significantly better clinical outcomes, including longer PFS and OS, compared with patients not receiving MTB recommended regimens. ⁶⁶

Treating patients based upon a classic drug-centric model that relies on finding commonalities between patients, initially the organ of origin of the tumor and more recently a common genomic alteration, may not be ideal if each patient bears a cancer that is both molecularly complex and unique. Moreover, within the same patient metastatic cancers are akin to “malignant snowflakes” in that they are distinct from each other and each harbors remarkable molecular complexity. ⁶⁷ Genomics has revealed the need for a patient-centric model of cancer treatment in which combinations of drugs are customized based on the molecular biology of each patient.

Gene expression assays and next-generation sequencing of somatic and germline genomes engender an ever-expanding list of prognostic and predictive factors, and each additional predictor elevates complexity and creates a web of interactions among emerging and established disease factors that is beyond comprehension using traditional approaches. ⁶⁸ To provide an accurate interpretation of the cancer status of an individual patient, clinicians must use all information available that will allow the complexity of the computational model to approach the complexity of the biological system. The only feasible means of synthesizing the magnitude and interdependence of such multimodel data is by the use of advanced AI. ⁶⁸

Critical to applying precision medicine is the concept that the right combination of drugs must be chosen for each patient and used at the right stage of the disease. Too often MTBs convene at a point when the cancer is refractory and has evolved in such a way that targeting actionable driver lesions might not derail the oncogenic process. ⁶⁹ Concurrent depletion of immune-suppressive factors and tumor intrinsic targeting and the use of combination therapies with distinct mechanisms of action pertinent to the patient's unique tumor biology are critical to prevent metastasis-associated resistance and promote long-term response. The principle is to treat early with first-line combinations of targeted, chemotherapeutic, hormonal agents, biologic agents, and immunotherapies customized and molecularly matched to N-of-One individual patients. ⁷⁰ MTB matching patients to therapies in first-line N-of-One in advanced cancers improved PFS and OS outcomes compared with patients who received treatment determined by their physician. ⁷⁰

Recent review of MTB experiences highlights the difficulty of estimating how many cancer patients really receive the proposed targeted therapy and the proportion of patients who actually benefit from performing an MTB-driven therapeutic indication. ⁷¹

AI and Physician Shared Care

AI technology currently outperforms humans in terms of image recognition, risk stratification, improved processing, and round-the-clock assistance with data collection, collation, and analysis. Potentially, AI could analyze and integrate clinical information, radiomic, immune-histopathologic, and genetic data, and predict response to proposed treatment of individual cancer patients. While AI cannot substitute for higher level interactions, nuanced critical thinking, or the ambiguities of communication of uncertainty, this amoral, impersonal pitiless technology may, paradoxically, enhance the personal empathic doctor–patient relationship and underpin shared care between oncologist physician and their patient. In shared-care decision-making the physician provides the medical facts and treatment options, patients provide their values, and together they form a plan that best matches the facts to the values.

The focus of MTB management is to make individualized treatment decisions for each patient based upon their clinical presentation. Although treatment plans are patient specific this approach may not be patient centered. Instead, it may present a barrier to effective shared decision-making as patient's values and preferences are not reliably acknowledged during MTB treatment discussion; this has been described as “health care in absentia” and may give rise to discordance between MTB recommendations and treatment implementation. ⁷²

It is conceded that, for now, the MTB process offers the only practical approach to state-of-the-art cancer care. However, the rapidity with which more effective innovative cutting-edge molecular genomic and immunologic targeted therapies are becoming available, and the advent of AI to manage real-world data in real time, behooves us to contemplate a fundamental change to our current MTB paradigm of cancer management.

Conceivably, an individual oncologist could use a trained AI system to perform all the current functions of a MTB in respect of data collection, collation, and analysis and even predict the likely efficiency of any eligible combination molecular precision therapy for the N-of-One patient. All of these AI-generated data and personalized management options could then be considered by the oncologist in one-on-one discussion with the patient. This personal communication could be informed and facilitated by the presence of a Chatbot natural language processing unit, now available with current internet access, functioning in real-time in any language or style convenient for rapid response to collectively based questions which may be beyond the expertise of the individual oncologist. ⁷³

It is important to recognize and to acknowledge the limitations and potential dangers of AI, which seems destined to surpass human capacity, not only in chess and Go but also in the game of life. Philosophically AI could conceivably attain ‘technological singularity’, which is a hypothetical point in time at which AI with human-level abilities irreversibly changes human civilization. We would then become inseparable from machines, which themselves evolve and redesign themselves at an ever-increasing rate with which our limited slow human biological evolution could not compete.

Regulatory authorities are scrambling to attempt to ensure that AI development is secure, trustworthy, and ethical. Legal and ethical challenges include data content of a chatbot, cybersecurity, data use, privacy and integration, patient safety, and trust and transparency between all participants. The European Union AI Act is designing laws based upon a need for explicability and a requirement for high quality data, documentation and traceability, transparency, and human oversight of accuracy and robustness. ⁷⁴

The recommendation for medical application is exploitation of the current capability of AI for data collection, collation, and analysis, while the potential interpretive and prognostic roles should always remain the responsibility of the treating oncologist. Practical experience, empathy, and interpersonal skills are not easily replicated by AI, and the ability to handle nuanced and complex medical situations in real world settings will remain the prerogative of the human doctor.

ChatGPT is a disruptive technology with the potential to fundamentally change how we interact with technology and perhaps to revolutionize the way medical professionals engage with patients. ⁷⁵ The traditional doctor–patient relationship could paradoxically be reinstated, through the intervention of Chatbot assistance, and, in contrast with MTB discussions, the intimate personal setting would invite exploration of spiritual and cultural values and expression of hitherto unspoken fears and heartfelt wishes of the cancer patient. Such concerns include the “financially toxic cocktail” of cancer treatment ⁷⁶ and the effect on the patient's family of such issues that are outside the purview of an MTB and which have a major effect on QOL.

Engagement expresses the commitment of the treating doctor to include patients in discussions about their care as individuals and in the coproduction of service plans. This shared-care is based upon trust and fosters empowerment of patients to gain control over their own lives and increases their capacity to act on issues that they themselves define as important to their QOL. ¹⁰ Given that a Chatbot may be part of the conversation, to what extent can the user “trust” Generative Pre-trained Transformer 4. ⁷⁷ Such conjecture is further complicated by the demonstration that the domains of trust in humans and in AI are not related with each other. ⁷⁸

The comprehensive recording of patient data within the clinical context of regular oncologist–patient consultations, particularly medical decision-making and outcomes, is now under development, using the advanced data processing capacity of AI, to introduce a novel clinical trial construct. The master observational trial hybridizes the power of molecularly based master interventional protocols with the breadth of real-world data. ⁷⁹ Prospective observational recording of biomarker and outcome data informing precision cancer therapy, reported by a vast network of individual treating oncologists, compatible with a master protocol, could address the inherent problem of interventional trials which usually focus on narrow molecular signatures and only one line of therapy, which is “akin to finding a single piece of a large jigsaw puzzle whereas personalized medicine represents the entire completed puzzle.” ⁷⁹

Another way of visualizing the big picture approach of AI in oncology is to image each patient's health as a digital photograph. Medicine attempts to decipher this image on a pixel-by-pixel basis, using studies based upon explicable pathophysiology precisely to elucidate individual pixels scattered throughout. Meanwhile, AI looks at the whole picture, starting as a poorly focused approximation but with increasing definition as improvements in databases, algorithms, and computational power continue. ⁶⁸

In nuclear medicine oncology, the essence of cancer theranostics in the real world has long been the practice of truly personalized precision medicine, to confer clinical benefit and enhance QOL, as evaluated through the perspective of the individual N-of-One patient. ⁸⁰ The benefits of theranostic targeted molecular therapy of specific cancers have been, hitherto, generally ignored, or undervalued, by MTBs. However, it is becoming apparent that the worm has turned. The increasing complexity of genome-informed molecular cancer management, necessitating AI assistance, may encourage a more personal, truly patient-centered, holistic approach to treatment across all oncology therapeutic modalities. The nuclear physician is well-positioned to take a leading role in this revolution in cancer care in which the modern technology of AI provides the power to recreate and amplify the doctor–patient relationship.

Conclusions

The continuing viability of MTBs is threatened by overwhelming patient demand, and the information explosion of molecular omics, compounded by the impossibility of performing RCTs individually for all biomarker-treatment-cancer type combinations, including chemo- and immune-therapy. The intrinsic complexity of biomedical text and vocabulary now requires novel approaches and infrastructure, such as those which may be afforded by AI, including natural language processing to bio-curate clinically relevant information on drugs, genes, diseases, and therapeutic opportunities. It is becoming impractical for members of MTBs to keep up with the rapidly growing volume and breadth of knowledge to reach timely and appropriate personalized cancer management decisions for each patient.

Given the advent of new generation AI systems to collect, collate, analyze, and interpret data from an individual patient medical record, omic profile, and response to past medication, the data with individualized prognostication and probability of response may potentially be communicated, through Chatbot natural language processing, directly to the oncologist, in the presence of the patient. The physician is thus empowered to evaluate the therapeutic options, to discuss them in real-time face-to-face with the patient, and to provide truly personal holistic cancer care. In the future MTB could be freed from taking responsibility for individual cancer management decisions to meet inevitably stressful time deadlines. The multidisciplinary specialists could then convene in a collegiate forum to discuss and review selected illustrative cases retrospectively, to evaluate the efficiency of evolving-omic-guided molecular targeting therapies informed by real-world follow-up. In particular, they could then determine actual clinical benefit, and cost, of novel treatments of cancer relevant to their own clinical practice.