Cancer is widely recognised as a human malady that embodies all elements of the triad of disease, illness and sickness; causing physiological malfunction, severe subjective symptoms and distress, and associated with certain specific social claims and obligations. The word alone spurs discomfort and fear, and is often shadowed by associations of struggle, pain, suffering and death (Vrinten et al. 2017). Historically these associations are both well placed and justified. The term cancer, originating from the Greek word “karkinos”, meaning crab, well captures with what malicious and ruthless force a malignant tumour can invade and spread its claws in a destructive and merciless manner throughout the human body. Until very recently the only form of cancer recognized in clinical medicine resembled this very image. Paired with mutilating surgery and toxic medical treatment terror has been considered a very suiting response to cancer. Currently, however, there are indications that the content and organisation of the category of cancer as a pathological process is rapidly evolving. We are presently in an era where a new form of medical practice is taking shape and gaining ground; precision medicine. Founded on the historical relationship between improved pathophysiological understanding and the subsequent development of efficient medical solutions this emerging medical approach proposes to change the way we approximate cancer as a disease process on several dimensions. Based on increased precision and resolution the hypothesis is that individual classification of cancer at a molecular level, rather than macroscopically and morphologically, as is the tradition, will promote improved medical management and result in reduced cancer related morbidity and mortality (Ashley 2016). As the study and clinical management of cancer steadily distances itself from the form and shape cancer takes on when it manifests itself clinically new questions and challenges are, however, bound to emerge. In this essay I would like to reflect upon some of the novel surfaces of friction that are currently materializing as the boundaries of cancer are set in motion by the shift in dimensionality that precision medicine promotes; moving from the macroscopic to the molecular level of understanding cancer, and from the study of humans to the study single individuals. To illustrate the complexity and intricacy of some of the issues that arise I will base the discourse on two interconnected narratives exemplifying the transition; the historical storyline about the coming of a molecularly defined subgroup of acute myeloid leukaemia (AML), and the story of the biological evolution from a normal hematopoietic progenitor cell to an AML cell. Based on these two stories I will proceed to discuss how the molecular level of resolution currently challenges not only how we academically and scientifically go about defining cancer the disease, but ultimately how this change in resolution also implies vast clinical and social consequences, correspondingly influencing both the illness and sickness aspects of cancer. As scholars of medical philosophy has not yet resolved in agreement on the central epistemological and ontological questions about the nature of disease and disease classification I will not dare to take a definite position. I rather try to reflect on the phenomena of cancer from the position of the simple clinician and translational cancer researcher that I am; as a product of biology, but always framed and managed by humans, and therefore essentially shaded by both aspects of underlying nature and normativity.

Leukaemia was first described as a distinct clinical entity in the midst of the nineteenth century. Initially the disease was recognised based on symptoms, distinct clinical observations and autopsy findings. Based on this level of resolution the first division of leukaemia into two subgroups shortly followed; an acute, storming and rapidly lethal form of the disease, contrasted to a more indolent chronic form of leukaemia. Only when microscopic assessment of the blood was available by the midst of the century did one approximate the nature of leukaemia as a neoplastic condition originating in the hematopoietic system and as the microscopic method became more refined the understanding of leukaemia gradually improved (Kampen 2012). Based on morphological distinctions leukaemia was subsequently subjected to an additional stratification; leukaemia originating from cells committed to the myeloid lineage of the hematopoietic system opposed to leukaemia deriving from the lymphatic cell lineages. Founded on morphological and cytochemical attributes further sub-classifications followed, and by 1976 three groups of acute lymphoblastic leukaemia and six distinct morphological subgroups of acute myeloid leukaemia were proposed, illustrating the gradual compartmentalization of the disease (Bennett et al. 1976). Contemporarily, the clonal origin of leukaemia was gradually uncovered, and the relationship between the disease phenotype, the cell of origin and genetic damage was gradually revealed (Nowell 1976). This resulted in a steady shift from a descriptive morphologically based system for organising the different sorts of leukaemia to a more pathophysiological and functionally founded organisational structure, frequently based on the molecular level of genetic aberrations. Initially cytogenetic features were the main focus, but as methods for molecular genetics became more refined and available also mutations have gradually found their place in the organisational system (Bennett 2000; Vardiman et al. 2009; Arber et al. 2016).

A molecular feature that has been devoted immense attention in AML from the very beginning of this shift towards molecular pathophysiological characterisation of cancer is the protein FLT3. The gene coding for this protein was identified as recurrently mutated in AML already in 1996, and the presence of mutations within this gene has consequently been linked to both phenotypical and functional properties. The presence of FLT3 mutations has been demonstrated to be statistically associated with features like age and gender, and clinical characteristics like leucocytosis and high bone marrow blast counts at time of diagnosis. The group has further been correlated with cytomorphological features, cytogenetics and molecular genetics, as well as with more functional properties like expression of certain immuno-phenotypical markers and intracellular signalling patterns. Ultimately the mutation has been repetitively coupled with patient outcome, predicting high likelihood of poor response to treatment, high relapse rate and inferior overall survival. Based on these findings and bound together by this shared pathophysiological feature the group has gradually been thought of as a confined subgroup of AML, and has consequently been treated as a distinct biological entity in preclinical and clinical research as well as in clinical practice (Lagunas-Rangel and Chavez-Valencia 2017).

As mentioned in the introduction, the leading clinical motivation for molecular characterisation of cancer and the ultimate validation of such an endeavour is closely linked to the utility of the approach. Within the field of haemato-oncology this might be best illustrated by the discovery of the Philadelphia chromosome, and the BCR-ABL oncoprotein in chronic myeloid leukaemia (CML). The molecular framing of this disease resulted in the subsequent development of pharmaceutical agents like Imatinib (Gleevec), specifically targeting the oncoprotein, resulting in remarkable therapeutic advances for this patient group (Deininger et al. 2005). Descriptively and functionally FLT3 mutated AML shares resemblance to BCR-ABL positive CML. As myeloid malignancies AML and CML share very similar cells of origin, and they are both seemingly driven by genetic aberrations resulting in an oncoprotein in the form of a constitutively active tyrosine kinase, resulting in a comparable proliferative advantage. Encouraged by the success of BCR-ABL targeted therapy FLT3 very early singled out as an attractive therapeutic target in AML. During the last two decades considerable effort has been devoted to developing therapeutic agents that would selectively benefit this patient group through specific inhibition of FLT3. 15 years after the initiation of the first clinical trials exploring the benefit of such FLT3-targeted therapy only very recently a few trials have demonstrated modest clinical benefit (Stone et al. 2017; Cortes et al. 2019; Perl et al. 2019). The fact that significant clinical improvements for this patient group still seems far out of reach is discouraging, and many possible explanations have been put forward attempting to explain the limited responses (Engen et al. 2014). At the core however, the lack of substantial achievements may possibly indicate fundamental limitations in the current labelling of FLT3 mutated AML. Accumulating data derived from the past 20 years is gradually exposing the heterogeneous nature within this confined group. The summation of temporal, spatial, multidimensional and high-resolution analysis of this group has revealed vast inter- and intra-individual heterogeneity. The gene can be damaged in multiple ways; most frequent by internal tandem duplications of varying lengths and motifs, followed by point mutations and occasional insertions or deletions. Uniparental disomy of the part of the chromosome entailing the mutated gene is also a common aberration, resulting in homozygous mutations and loss of the wild type allele. All the variants are validated as functionally significant, although with varying implications. The fraction of the leukemic disease that entails a FLT3 mutation is in addition highly variable and dynamic, ranging from a diminishing portion of the leukaemia cells to defining the entire tumour cell population. A large portion of the patients actually has several sub-clones, characterised by different distinct FLT3 alterations. The mutations can even occur or disappear through a single clinical disease course. Several of these features have further been linked to disease specific outcome; the length of the mutation, portion of FLT3 mutated cells, the insertion site of the length mutation, and the amino acid sequence of the duplicated region are just some of the features various investigators have attributed functional significance with prognostic implications (Lagunas-Rangel and Chavez-Valencia 2017).

While the inter-individual heterogeneity of FLT3 mutations makes the group hard to study, it might essentially be the intra-individual heterogeneity and kinetics that generates the greatest challenges in the practical assessment and categorisation of this patient group. What this intra-tumour variability implies, however, is that alterations in the FLT3 gene most frequently represent a late event in the process of leukemogenesis, and this is essentially where the great value in the data positions. When the individual findings are aligned and considered together they add up like pieces in a puzzle, revealing truths about the second story I now want to tell – the tale of the origin and evolution of leukaemia; the transition from a healthy cooperative and obedient hematopoietic progenitor cell to an insensitive and anarchistic leukaemia cell, and the story starts long before the FLT3 mutation enters the narrative.

Cancer is often described as characterised by their monoclonal origin, and this implies also for FLT3 mutated AML. The leading hypothesis, supported by strong emerging evidence, suggests that preceding the FLT3 mutation, maybe by as much as decades in some patients, an initial molecular alteration occurred in a long-lived hematopoietic stem or progenitor cell. This alteration resulted in functional changes in the cell, providing a survival advantage, securing that as time passed by the decedents of this very first cell survived and/or multiplied at a higher rate than the other progenitor cells. This ultimately generated a pool of cells characterised by this common alteration, confining a condition we now call clonal haematopoiesis of indeterminate potential (CHIP) (Genovese et al. 2014). As occurrence of genetic damage and epigenetic modifications are frequent events, at a point in time an additional alteration occurred, providing an added advantage to one of the cells, and then this cell again fostered a group of decedents with shared properties. This process continued in a branched manner, gradually generating a vastly heterogeneous pool of cells with varying degree of deviant behaviour, although with certain remaining similarities. At a certain point one of these cells gained a property that influenced also the interaction with surrounding cells and tissue, and a situation where normal hematopoietic function was affected arose, resulting in what we might define as a pre-leukemic condition (Shlush et al. 2014). Eventually the sufficient damage befell one pre-leukemic cell resulting in the development of the full blown leukemic phenotype. Within the group we currently characterise as FLT3 mutated AML potentially, and in some cases probably, it is the acquisition of exactly the FLT3 mutation that generates the leukemic phenotype and initiates the clinical presentation of the disease. The augmented proliferative drive the damaged FLT3 protein adds to the already damaged cell could be the sufficient addition of disruption needed for the cell to divide at such a rate that it and its descendants manages to overtake the bone marrow environment, effectively causing disruption of normal hematopoietic functioning and resulting in the presentation of symptoms, and clinical findings.

The two stories presented, the first about the formation of FLT3 mutated AML as a diagnostic unit and the second about the evolution of FLT3 mutated AML as a biological entity, can be condensed to illustrate two central challenges; (1) The soundness of FLT3 as a marker for the transition from a non-disease to a disease and (2) the legitimacy of FLT3 as marker of a specific confined disease. The generalisation of these two questions transfer us towards a major challenge in precision oncology; if and how a defined molecular feature or event can reliably and truthfully delineate something healthy from something diseased, or serve as robust foundation for dividing and stratifying specific diseases. So after the exercise of contracting down from the level of illness and clinical symptoms to the resolution of single molecules in leukaemia I would like to reverse and expand the perspective, attempting to demonstrate the relevance of this example when considering the challenges within the overarching category of cancer.

Of both intellectual and clinical interest is the question of whether the acquisition of a specific molecular alteration can delineate the point in carcinogenesis when a cell or a group of cells transit from being normal or healthy to becoming a disease or a condition that can be defined as a pathological process. Returning to the example of leukemogenesis and whether the attainment of a FLT3 mutation can serve as marker for the conversion from non-leukaemia to leukaemia it seems from a naturalistic perspective plausible that the FLT3 mutation may well denote a cell fraction characterised by the fully developed leukemic phenotype, at least on a single cell level. The clinical relevant disease however is frequently demonstrated to entail additional complexity. Experience treating FLT3 mutated leukaemia has shown that although the addition of a FLT3 mutation may be the sufficient element founding a confined unit of cells responsible for the promotion of the clinically début of leukaemia, it is clearly not necessary, as there are other mechanisms that can produce a very similar phenotype. After successful induction therapy in a FLT3 mutated patient the leukaemia can recur as a FLT3 wild type disease, or even being characterised by an unrelated novel FLT3 mutation. The existence of AMLs characterised by only a small fraction of FLT3 mutated cells indicates the same conclusion, implying the presence of alternate leukemic drivers in the remaining FLT3 wild type blast population. The presence of several distinct FLT3 mutations concurrently within the same AML provides further proof; as such mutational patterns are shown to define different distinct cellular populations with discrete functional properties. This heterogeneous pattern is certainly not valid just in FLT3 mutated AML but also in other cancers where it has been shown that although an individual cancer is usually characterised by a common driver mutation, distinct cellular subsets are characterised by additional but discrete driver mutations, and that the composition is dynamic. The spatial heterogeneity is particularly striking with evidence suggesting that metastatic disease often derives from ancestral cells, preceding the dominating genotype of the primary tumour (Gerlinger et al. 2012; Johnson et al. 2014; Gibson et al. 2016). The parallel co-occurrence of several distinct disease characterising driver mutations challenge the core of precision oncology and raises the question of how these situations should be interpreted and managed.

The conclusion one might derive from this is that if FLT3 denotes the transition on a single cell level, then most patients with AML have many different and distinct parallel leukaemias. In addition most AML patients probably have several additional conditions, including several distinct pools of clonal hematopoietic cells and multiple pools of pre-leukemic cells. The same situation would consequently apply for most cancers at the time when they revile their malignant nature. Precision-wise one may therefore consider the situation as managing several distinct diseases and conditions within single individuals, some times not even separated by time, but simply by molecular characteristics resulting in distinct functional properties. The fundamental challenge here is that dependent on the strength of our magnifying glass all cancer cells could in nature be unique, both in descriptive and in functional terms. When considering the process of carcinogenesis in mechanistic terms, as the result of numerous sequential molecular alterations, happening one at a time in parallel in every single cell; one acetylation here, one posttranslational modification there, one double stranded DNA brake here, and one phosphorylation way over there, it is evident that no cancer cells are identical in molecular terms. Supportive evidence of the functional consequences of this is derives from studies of relapsed AML patients, after treatment with selective FLT3 inhibitors. These therapeutic agents pose a very specific selection pressure on the leukemic cells, and the recurring result is emergence of a polyclonal pattern of novel FLT3 mutations conferring resistance to tyrosine kinase inhibitor treatment. This indicates a narrow selection of cells, otherwise not detectable, that likely under no other condition would be selected for or would stand out as functionally different to the bulk disease, except exactly under these very specific conditions (Man et al. 2012; Smith et al. 2012, 2017; Baker et al. 2013).

The thought experiment above leads to the possible conclusion that every individual cancer cell may ultimately represent a discrete functional unit (at least potentially – dependent on applied selection pressure) and may thereby even be considered individual cancers in biological terms. However, at this level of resolution the clear-cut patterns and delineations of what can be considered normal or deviant becomes difficult to grasp, and the close relationship between the level of dimensionality and what can be considered normal or anomalous becomes very clear. Recalling the transitions of dimensionality illustrated in the stories told about FLT3 mutated AML this becomes very clear. In developed countries cancer is a leading cause of death and a very common disease affecting up to one third of the population in a lifetime perspective (McGuire 2016). Within the category of cancer haematological malignancies are however rather rare, and the lifetime likelihood of being diagnosed with AML is very low. Incidence is related to age, and presentation of AML early in life is extremely uncommon. Even more improbable is the development of FLT3 mutated AML. Increasing the sensitivity when assessing FLT3 mutated AML patients I have tried to demonstrate that molecularly they are all unique; in fact all of their cancer cells might ultimately represent distinct biological entities, as derived from the conclusion above. Reversing the reasoning the same challenge emerges; damaged hematopoietic progenitor cells with future leukemic potential are most likely almost ubiquities and if we just use a sensitive method enough we would probably show that we all possess such cells occasionally (Young et al. 2016). Enrichment of such cells as to fit the current definition of CHIP is however slightly more rare (Genovese et al. 2014), and the presence of pre-leukemic cells posing a danger to normal hematopoietic homeostasis is likely much more infrequent. Overt leukaemia remains very rare. The same reasoning applies when considering changes in function – are we to consider fluctuations in in function on the level of the individual, of the organ, of the tissue or the individual cells? The chosen dimension of investigation seems to strongly influence the statistics and thereby conclusions produced, and the question of what the appropriate level of inquiry should be seems of outmost importance. How should we understand cancer and the dangers of cancer under such varying perspectives? To take this analysis to its outer limits, what it implies is that (1) we all have cancer if we just look closely enough, and (2) all these cancers will be singular diseases that on a molecular level only can be grouped together by the use of simplifications.

From a clinical perspective this conclusion may not intuitively seem very useful, so moving slightly from the discourse of what can be considered real to what is practically relevant within the realm of applied medicine some significant practical questions emerge. From a clinical and therapeutic perspective it is not just about what is and what is not, but maybe more importantly about what could and should be done. The principal goal of precision oncology is after all to improve patient care and outcome. Molecular markers are merely considered the tools required for achieving this. Although surrounded by significant hope and hype, the unsatisfactory results in the clinical management of FLT3 mutated AML (Prasad and Gale 2016) unfortunately mirrors the current status and accomplishments of precision oncology far better than the triumphant story of tyrosine kinase inhibitors and their achievements in the treatment of CML (Prasad et al. 2016). As I previously challenged the validity and utility of labelling of FLT3 mutated AML as a discrete diagnostic subgroup based on the lack of clinical improvements, one might question if a similar challenge may essentially be an underlying variable currently restricting the potential of precision oncology at large.

To start at the very practical oriented end the viability of precision medicine as project, both scientifically and clinically, is supported and dependent on powerful and high-resolution technological solutions, allowing uncoupling of the subjective and clinical findings of cancer traditionally causing illness, by focusing the gaze on ever- more narrowly defined composites and interactions of the biological processes involved in the development of cancer. An apparent challenge is nevertheless that the precision of the lines we draw is rigorously limited to the qualities of our scientific assays. In fact, examining the historical narrative of the compartimalisation of leukaemia it is from beginning to end largely framed and defined by available technology, ranging from the early development of the microscope to the recent technological advances of single cell next generation sequencing. Considering FLT3 mutated AML in specific it is clear that technological limitations shade the true incidence and prevalence of FLT3 mutations in AML. The characterisation of this gene in AML is routinely performed by polymerase chain reaction amplification, fragment analysis and Sanger sequencing. The conditions of the assay, ranging from the amount of DNA put into the reaction to the in silicointerpretation of the data ultimately influences the sensitivity of this analysis. The ratio between FLT3 mutated and non-mutated patients in AML it is therefor reflecting the chosen settings of the scientific analysis rather than the “true” naturally occurring ratio. While most articles discussing FLT3 mutated AML states that the mutation occurs in approximately one third of AML patients assessments by more sensitive methods have revealed that the portion of FLT3 mutated patients is substantially larger (Ottone et al. 2013). This ultimately limits the validity of the compartimalisation of subgroups of cancer based on the presence or absence of most molecular traits.

Moving from the use of molecular markers as delineation of important and relevant disease fractions to the use of molecular markers as indicators of disease from a clinical perspective, at least within the frames of precision medicine, the point demarking the transition from a state of subclinical disease to debut of clinical disease may essentially not be the point of greatest interest. While only the last molecular steps in the process of carcinogenesis may ultimately be good markers and predictors for clinically relevant disease, they may conversely be of little use as predictive markers, they may be unfit as efficient therapeutic targets and they may serve as poor markers for therapy surveillance and disease recurrence. This is at least indicated from the experience of the study and treatment of FLT3 mutated AML.

Predicting the onset of cancer is a major focus in precision medicine; with the goal of preventing the development of clinical relevant disease. This in essence means that one is actively searching for occult pathological conditions not yet promoting symptoms in individuals that consider them selves as healthy. The fact that FLT3 mutations are late events in leukemogenesis makes them poor markers for prediction of future disease. It is rather the mutations characteristic of clonal haematopoiesis and pre-leukaemia that are feasible for early detection. Searching, identification and management or premalignant conditions does however come with some major pitfalls. What the story of the biological evolution of FLT3 mutated AML has demonstrated is that the earliest steps of leukemogenesis is only weakly related to the presentation of overt leukaemia in the future. Applying highly sensitive methods mutations defining CHIP, and accordingly also recurrently mutated in AML, has been demonstrated almost to be ubiquitously present in adult individuals (Young et al. 2016). We know that very few of these individuals will ever progress to develop any sort of clinical relevant haematological malignancy. Screening for and treatment of CHIP and pre-leukaemia therefor presupposes a willingness to withstand a substantial increase of total individuals diagnosed and treated for something that if left untouched would never develop into clinically relevant disease.

Moving from prediction to action the experience from targeted treatment, including FLT3 targeted therapy, indicates that therapeutic success, as in achievement of clinically meaningful responses, may necessitate a broader aim than targeting what is understood as the latest acquired molecular drivers of the disease. The entire pool of cells descending from the initial monoclonal origin may possibly be vital parts of the biological entity of the disease, and cancer cure may in biological terms signify that not only the various disease fractions but also the premalignant fractions of the disease must be managed to secure the prevention of disease recurrence. Many have suggested that earlier molecular events may serve as much more attractive markers, not only in prediction and prevention of cancer, but also as therapeutic targets and in surveillance of disease recurrence. Combination treatment or sequential therapeutic regiments are the alternate strategy, but how many targets that need to be considered to achieve therapeutic efficacy in relation to how many targets that can be managed without accumulating unacceptable toxicity is still unanswered questions. Importantly, however, we know from the study of minimal measurable disease in AML that remaining fractions of both clonal haematopoiesis and pre-leukaemia is frequent, and that although associated with variable risk of recurrence it is by far deterministic (Hirsch et al. 2017), bringing us back to the ambiguous nature of pathological conditions that are not causing clinical implications when assessed. Again, increasing therapeutic intensity based on traces of remaining cells characterised by their heritage as descendants of the first monoclonal cell of origin may signify significant overtreatment. A crucial question in the undertaking of predictive oncology – both predicting disease development and disease recurrence, is if the development of cancer causing clinical relevant disease will ever be possible to reliably foresee. Considering the level of randomness that seemingly is involved in the evolution of cancer there is a real possibility that the prediction of cancer forever will be associated with estimating risk rather that certainty. A deciding element in the way we end up managing early stage cancer and evidence of remaining non-disease promoting cells after therapy may therefor reside far beyond the limits of biology, but rather reflect our willingness and ability to manage risk.

In this essay I have attempted to illuminate some of the challenges that arise in the process of transforming the boundaries of the category of cancer in two central dimensions – from groups to individuals and from the macroscopic level to the molecular level. As the potency of our magnifying glass increase, and the shift towards the highest imaginable level of resolution advances the distinct diagnostic boxes are becoming increasingly refined and precise. The consequence seems to be that the validity of what is “normal” from a quantitative and statistical perspective seems to gradually dissolve. The heterogeneity of cancer is no longer limited to distinctions between cancers of the blood compared to cancers of the gut, or to distinctions between cancers driven by specific mutations. The variation is made visible, tangible and relevant down to the level of not only individuals but of single cancer cells. Cancer is, thus, not a static biological being; it is in essence evolution and thereby dynamic in character. The same kinetics, however, seem also to apply for cancer when considering its broader character; as the totality of cancer as a disease, illness and sickness. As a malignant tumor is known to spread throughout the human body cancer has similarly gained a pervasive grip on contemporary western society, where it plays a central role in the structuring of both healthcare politics, healthcare services, and the private healthcare and pharmaceutical industry. There is a multitude of individual stakeholders in this process, not only suffering, but also profiting on cancer, and ultimately where the lines are drawn around the category of cancer has potential pervasive cultural, political and economical consequences. The central question of this essay sounds: when the macroscopic boundaries of cancer the illness and disease dissolve and loose their relevance, where are the new borders to be drawn? The complexity of any relevant answer should probably entail elements not only on where but also how, why, and by whom these decisions are to be made. Independent of the origin of the applied boundaries of cancer however; as a naturally defined pathological process, confined by available technology, drawn as a result of social constructivism and normative choices, or all of the above, the category remains powerful not simply as a creature of nature but through its collective position in society. The expansion, reduction or stratification of the concept therefore has widespread effects outside the realm of biology. Already the meaning of cancer has changed in such a way that it does not always equal the devastating, destructive and mortal human malignancy that we often associate with the word, and the traditionally firm relationship between cancer as a disease, illness and sickness is consequently fading. If the goal of precision medicine is to be achieved – to reduce cancer related suffering, then as the biologically founded disease fraction of the category cancer expand, cancer as an illness and sickness must synchronously decrease. Ultimately the high level of resolution in the study and understanding of cancerchallenges not only how we traditionally go about classifying cancer, but even more notably how we consider it as a distinct and definite human malady and health threat. Together these elements indicate that the current transferral towards precision oncology is not only a practical and clinical endeavour but represents both epistemological and ontological challenges that need to be addressed.