Causal Histories of Psychological Factors and Cancer: From Psychosomatic Medicine to Neuroimmunomodulation Open Access

Some authors have argued that using our point of view of the present can be a fruitful starting point to look back on history under certain circumstances [1‒3]. Indeed, history, thus construed, can provide scientists and physicians with more context regarding their present hypotheses and methods and, hence, encourage more critical thinking regarding methodological and conceptual aspects of their work [3‒6]. Even though the history of ideas is often opposed to that of techniques and instruments, it is important to consider the articulation of both when addressing the history of life and health sciences. In this respect, and in spite of the opinion of many of today’s practitioners that the way they do science is the way it should always have been done, it is important to acknowledge that ideas about causality, experimentation, observation and objectivity, for example, have also changed over time [7].

Indeed, ideals for causal thinking and explanations in science and medicine have changed in important ways over the past 100 years. Regarding these changes, it seems safe to say that mechanistic causation has gained a prominent place in biology [8], while in medicine mechanistic explanations have been complemented by a more statistical view of causation [9]. Instead of a “broad-stroke” history of causality in biology and medicine, the emphasis here will be on cancer because several types of causal explanations have been proposed regarding cancer over the past decades [10‒12]. During the last quarter of the 20th and continuing into the 21st century, the dominant somatic mutation etiological model of cancer stipulates that this disease is due to the accumulation of DNA mutations in a somatic cell resulting in unusual cell proliferation that is not sufficiently noticed by the immune system [10, 13, 14]. According to the minority tissue organization field theory of cancer, carcinogenesis results primarily from a problem of tissue organization [10, 13, 14], the surveillance of which has recently also been proposed to involve the immune system [15]. So, regardless of the etiological theory of cancer that one adheres to or the functional role that one ascribes to the immune system, the latter will be considered to play an important role in cancer progression. Even though several hormones and neuromodulators can influence tumor growth independently from immune mediators, it is therefore not surprising that, in the past 50 years, many studies that addressed the links between psychological factors and cancer have focused on immune system components. Within this context, the focus will be on how interdisciplinary fields, such as psychoneuroimmunology, have considered causal links between psychological factors and cancer.

The choice to focus on the potential causal links between psychological factors and cancer is not an arbitrary one but corresponds to still-open questions that have repeatedly been asked by physicians and scientists in the history of research on cancer. These open questions concern the psychological causes of cancer [16‒20], as well as the psychological consequences of cancer [21, 22]. The aim here will be to provide some insights into different causal notions that philosophers of medicine and science have distinguished to help clarify terms of the debate on the links between psychological factors and cancer.

The first section of this paper will introduce the importance of scientific pluralism when it comes to methods, evidence and concepts. The second section will introduce some causal accounts with regard to the links between psychological factors and cancer that have been developed over the past 100 years. These accounts will then be deepened in sections 3–5 with (1) a psychogenic interpretation of psychosomatic medicine that was influential from the early 1940s until the 1970s, (2) a discussion of stress mechanisms in cancer, and (3) a coverage of “difference-making” accounts specifying links between adversity and cancer, both of which gained traction from the early 1980s onward. The final section will then consider the possibility of a synthesis between these perspectives before concluding.

In several domains of science and medicine, a pluralism of methods is promoted with, for example, the idea that using various methods, data sets and analyses as well as different notions of causation leads to more robust conclusions [23, 24]. Such stances are, in part, motivated by the idea that, in philosophy of science and medicine, different ideas and explanations of causation have emerged and encouraged a “causal mosaic” vision in which one can place different causal tiles to obtain images that are fitting for one’s research projects [24].

One distinction among types of causality that is especially relevant for biomedicine is that of singular versus general causation. Biologists are typically most interested in general causation to explain how living phenomena work in general using statistical techniques, whereas a physician needs to determine what particular disease process accounts for the symptoms of an individual patient. Accordingly, a so-called “agency causation” account could allow consideration of subjective patient reports among the causal factors of interest. A second relevant distinction, both for medicine and biology, is between so-called “difference-making” or “dependence” causal theories that determine if a candidate cause A makes a difference to effect B and “production” causal theories that establish (a mechanism for) how A brings about B [24]. A final distinction of interest is that between the causation brought by a whole (molar causation) and that produced by parts of a whole (molecular causation [24]). This is particularly important to keep in mind when mental causal contributions to somatic pathological processes like tumor growth (downward, molar, or top-down causation) are proposed. Applying this to the question of psychological factors and cancer today, it seems that it is widely accepted that lower level molecular mechanisms associated with neoplastic growth can causally explain psychological changes in individuals with cancer (bottom-up, molecular, or upward causation).

Not surprisingly then, and to get back to the question of the necessity to propose several causal tiles to create a fruitful causal mosaic, it would be useful to obtain indications of both “difference-making” and “production” accounts of causation following the general idea that a stronger case for causation can be made if one can fulfill the criteria of several accounts of causation [24‒26]. Accordingly, the types of evidence that one looks for depend on the kind of causal claim that one would like to make. In a “difference-making” account the aim is typically to find evidence indicating if factor A can be considered a cause of phenomenon B or not, whereas in a “production” account, the objective is often to obtain evidence indicating how a lower level mechanism in A can bring about B.

A first type of pluralism that should be welcomed is therefore methodological, in the sense that many different methods and approaches can be useful in addressing causal relationships [23]. This, in turn, leads to a pluralism of evidence, since different sources of evidence are typically required to back up causal claims. From this, forms of causal pluralism can be derived that concern either our ways of knowing and explaining the world or claims about the world [24].

Triangulation can be seen as an active strategy to maintain and foster pluralism and can be defined as “the combination of at least two or more theoretical perspectives, methodological approaches, data sources, investigators, or data analysis methods” [23]. Through the use of triangulation, one can counterbalance the limits and biases associated with one single methodological strategy. This, in turn, increases the robustness of findings and one’s ability to interpret them. In the experimental sciences in particular, it has been proposed that methodological triangulation can result in unique and useful information that could not have been obtained following one single approach [23]. For example, the outcomes of recent intervention studies with optogenetic approaches can be compared to those of older pharmacological inhibition strategies in similar conditions of interest. It is important to note here that while the triangulation of findings obtained with optogenetic or pharmacological intervention strategies increases the robustness of a difference-making account of causation, these strategies can both be considered as variants of a “difference-making” account. In contrast, a causal mosaic typically involves several accounts of causation.

Other forms of scientific pluralism, beyond mere tolerance of other approaches, include integration or co-optation of elements of different fields or disciplines and even productive competition [27]. It is, however, important to keep in mind that there is not a single way to combine different approaches, be that for triangulation, integration or causal mosaic. Instead, it has been proposed that new attempts to build fruitful causal mosaics can be undertaken for every research topic [24]. This is what we set out to do here by considering different historical periods over the past 100 years or so, with regards to the causal relationship between cancer and psychological factors.

In the early 20th century when cancers became more widespread, but with their causes being largely unknown, some “blaming the victim”-kind of explanations invoking psychological factors emerged alongside more cellular-based explanations of cancer [28]. The proposed psychological causes of cancer were in line with older ideas, but were integrated in the broader context of psychosomatic medicine in the 1940s and 1950s [28]. Psycho-oncology is a line of research and care that emerged in the 2nd half of the 20th century that focused on the psychological responses of patients after they have received a cancer diagnosis [29, 30]. Post-WWII, causal links to cancer, based on epidemiological approaches, started to be put forward, most famously between cigarette smoking and lung cancer [31, 32]. In the last quarter of the 20th century, the claims of psychosomatic medicine started to be investigated more experimentally with the understanding that psychological stress can influence the course of cancer and that cancer and its treatments cause stress. Such research has been getting some attention from major cancer journals but has mostly ended up in specialized journals, such as Brain Behavior and Immunity, founded in 1987 by the Psychoneuroimmunology Research Society, and Neuroimmunmodulation founded in 1994 by researchers Samuel McCann and James Lipton. Since the immune system is considered to play an important role in cancer progression (see introduction), these specialized journals emphasize interactions among the nervous, endocrine, and immune systems to explain links between somatic disease, including cancer, and stress.

In the 1930s, some institutional support emerged in the USA in favor of the idea that psychological factors could cause somatic diseases, which then contributed to the foundation of the journal Psychosomatic Medicine in 1939. Even though, cancer was not part of the typical psychosomatic diseases of the time, many of the publications invoking psychogenic causation of cancer would later be published in that journal [33, 34]. These early reports led George Engel to wonder “(w)hy are certain patients approached with the bias that psychological processes are involved in the genesis of the disorder, while in others the psychological processes are considered reactions to an underlying somatic process?” [35]. His explanation included the idea that “‘bad’ emotions causing the disorder” based on the fact that emotions are associated with physiological changes, which, in turn led to “equat(ing) ‘psychosomatic’ with ‘psychogenic’” [35]. However, adopting such a psychogenesis stance, in turn, can expose one to a selection bias when studying the links between psychological factors and somatic disease. Engel, instead, proposed to adopt a multifactorial approach to cancer development and to strive for better psychological care of suffering associated with cancer [35].

As a way to assess how Engel’s constructive criticism landed, it can be insightful to have a look at the proceedings of the New York Academy of Sciences-hosted conference on the “Psychophysiological aspects of cancer” published in 1966. In his preface, the then president of the Academy announced that it is no question that the “‘set of the spirit’ can alter the course of …life-threatening disease” and proposed that “psychophysiologic factors [also] may determine the onset of cancer” [36]. The resulting volume of proceedings contained nine articles on possible psychogenic causation of and contribution to cancer. However, none of these proposed any strong conclusion.

Interestingly, among these papers, some wished that “neuroendocrinological and hormonal mechanisms which mediate between experience and biological process, most likely soon will make it feasible to study the biochemical and physiological processes which intervene between the narcissistic and regressive discharge of psychological drive and the creation or facilitation of neoplasm” [37]. Such hope was further reinforced by the claim that “neurological changes precede the onset of cancer and therefore cannot be attributed to wasting [38]. Others, instead promoted a more epidemiological approach and compared a cohort of over 2,500 individuals with characterized personalities, in terms of cancer incidence at a 10-year intervals to suggest that “Among women there is a much higher proportion of cancer cases among the substable (personalities) than among the controls (p < 0.005)” [39].

Interestingly, the volume of proceedings of the conference on the “Psychophysiological aspects of cancer” also contained six articles discussing the effects of cancer on psychological factors and ways to promote patients’ well-being. Furthermore, three articles in the volume focused on theoretical considerations regarding the relationships between cancer and psychological factors. One of these reminded that, in spite of the focus of the conference, it should not lead to the idea that “the psychological approach is thought to be the only, or even the most important, approach to [cancer], whether we are dealing with etiology or management” [40, 41]. Thus, the hope was also expressed that: “We are no longer concerned with such either-or single-factor polarities but currently attempt to map out the widest possible ranges of conditions, all of which in some way and at some time, seem to be implicated in a dynamic chain of causes and effects” [40]. Similarly, in the early 1970s, psychosomatic medicine was presented as a perspective that favored a multifactorial etiology of cancer and was thus complementary to a reductionistic “‘germ-theory’ view of disease” [42].

As indicated above, even though cancer was not part of the typical psychosomatic diseases, many of the publications invoking psychogenic causation of cancer were published in the journal Psychosomatic Medicine. We, therefore, propose to start the historical period during which research was driven by the idea that specific psychological factors could give rise to specific cancers in 1954 with the publication of a “special article” on this question in that journal. It is important however to point out that psychosomatic medicine has undergone several changes since. In particular, authors within the field have attempted to downplay the exclusive causal role of psychological factors and favor a more holistic vision compatible with multicausality [43]. In addition, it can be argued that psychosomatic medicine has contributed to the emergence of several contemporary fields, including psychoneuroendocrinology, psychoneuroimmunology, consultation-liaison psychiatry, behavioral medicine, health psychology, and quality-of-life research [44]. It is therefore not that surprising that at the beginning of the 21st century, psychosomatic medicine became a subspecialty of psychiatry after the recognition of “a high prevalence of psychiatric disorders … in patients with complex medical illnesses” [45]. Thus, 21st century psychosomatic medicine can be expected to focus more on quality-of-life questions in an approach that bears similarities to that of psycho-oncology. Unfortunately, and in spite of these evolutions, lay-audience-directed newspaper articles at the turn of the 20th century, continued to use the term psychosomatic to refer to something imaginary or of psychological origin [46].

Regardless of whether psychosomatic medicine is best conceived of as a discipline, a subfield of psychiatry or an approach, one could ask what its historical contribution has been. Besides putting forward a heavily criticized audacious hypothesis, according to which psychological factors can cause somatic disease, psychosomatic medicine can at least be credited for having put psychological factors on the agenda for investigation in biology and medicine. The idea that a person’s psychology can contribute to a somatic disease such as cancer bears some resemblance to so-called agency causation. In a broad philosophical perspective, “an event A is a cause of a distinct event B just in case bringing about the occurrence of A would be an effective means by which a free agent could bring about the occurrence of B” [47]. Adopting an embodied, enactive, embedded, and extended (4E) cognition perspective may be able to link the philosophical and biological accounts of causal agency. Indeed, within such a framework, agency has been proposed to result from loops of action and perception and to include top-down cognitive control of action, as well as functions of emotions [48, 49]. Thus envisioned, and if functions of emotions include bringing about physiological changes in the body, then agency causation can be considered to capture an important part of the strong thesis of initial psychosomatic medicine.

After the criticism of psychosomatic medicine for being too focused on individual cases and anecdotes [50] and the hope that “neuroendocrinological and hormonal mechanisms” will provide new insights into the relationships between psychological factors and cancer [37], it should not come as a surprise that a more mechanistically-orientated notion of causality started to become employed regarding the relationship between psychological factors and cancer. However, this seemed only possible after the notion of stress had been operationalized and shown to be associated with some bodily responses.

After discussing how physiological stressful situations are accompanied by hormonal secretions from the pituitary and adrenal (corticosteroids) and acknowledging the psychosomatic medicine framework, Hans Selye proposed that emotional stress, such as immobilization of animals, also give rise to the same bodily reactions [51]. These responses include hormonal secretions from the pituitary and adrenal glands as well as changes in the autonomic nervous system [51, 52]. Of note, the above-mentioned volume of conference proceedings on the “Psychophysiological aspects of cancer” also contained several articles discussing the links between stress and cancer in animals. These addressed the effects of stress on cancer and concluded, based on a review of the literature, that neuroendocrine mechanisms proposed by Selye could constitute mediating mechanisms [53‒55]. Overall, the findings available by the end of the 1960s were judged to justify the conclusion that “under certain circumstances responses to cancer (tumorigenesis and mortality rates) in animals can be influenced by experiential and/or environmental manipulations of the organism” [56].

Because, at the time, many cancers were thought to result from viral infection and because corticosteroids were known to be immunosuppressive, many considered the role of these hormones and immune responses in the effects of stress on cancer. Some clinicians rehearsed the idea that “personality factors, stress, and … the onset and course of cancer” are linked and proposed that this may be brought about by “stress-responsive adrenal cortical steroid hormones, which may be immunosuppressive” [57]. Another line of research took advantage of the long-standing consensus that Pavlovian conditioning of bodily responses did not require invoking subjective elements to explain the phenomenon of interest, even though it was clear that “something cognitive” had occurred. Interestingly, Pavlovian conditioning of leukocyte and antibody responses had already been shown in the early 20th century [58]. Later, In the 1970s, Robert Ader and Nicholas Cohen showed that the immunosuppressive effects of cyclosporin on antibody production can also be conditioned [59, 60]. Just like for the effects of stress, different mechanisms, including autonomic nervous system innervation of immune organs and neuroendocrine axes, were considered by the nascent field of psychoneuroimmunology to explain conditioning of immune responses [61, 62].

In the first edition of the monograph Psychoneuroimmunology, two chapters, one on humans, and another on animals, dealt with the links between psychological effects and cancer and discussed how catecholamines and corticosteroids, could influence tumor growth via their effects on the immune system [63]. Acknowledging the importance of epidemiological approaches (see next section) to identify stress as a risk factor for cancer progression, a review article proposed that this could be mediated through hypothalamo-pituitary-adrenal axis and sympathetic nervous system activation resulting in reduced immune responses and increased proliferation of oncogenic viruses and cancer cells [64]. Many publications have attempted to specify stress paradigms and to identify potential cellular and molecular mediators affecting cancer. Social defeat in rodents has thus been shown to favor the retention of tumor cells in known metastatic sites, such as the lung, in a beta-adrenergic receptor-dependent manner [65, 66]. Of note, a passive coping strategy seems to render animals more vulnerable to develop lung metastases [67]. It is important to remind that repeated social defeat results in activation of both the hypothalamo-pituitary-adrenal axis and of the sympathetic nervous system and alters immune cell production, status, and trafficking [68]. Furthermore, repeated social defeat is also accompanied by lung expression of adhesion molecules and increased metalloproteinase phosphorylation, which can enable immune cell evasion, but also favor tumor cell metastases [69].

Interestingly, surgery stress both in humans and animals has been found to promote cancer metastasis through the action of catecholamines, in part through their effects on immune cells [70, 71]. But besides their effects on immune cells, beta-adrenergic signaling also up-regulates genes that play a role in tumor progression and metastasis, while glucocorticoids increase the expression of genes that favor cancer cell survival and resistance to chemotherapy [72, 73]. The parasympathetic branch of the autonomic nervous system in the form of the vagus nerve has, instead, been shown to have protective effects against some metastases [74].

In the early 21st century, based on research showing that pro-inflammatory cytokines mediate changes in behavior in experimental models of infection, it was hypothesized that pro-inflammatory cytokines could be involved in the psychological effects of cancer and/or its treatments [75‒78]. Interestingly, women who experienced childhood adversity, such as abuse or neglect, reported greater fatigue and more depressive symptoms and showed lower natural killer cell activity and higher plasma interleukin-6 concentrations after breast cancer surgery compared to women who did not undergo childhood adversity [79]. These findings suggest that early-life stress can exacerbate some of the symptoms associated with cancer or its treatments.

We propose to start the still ongoing historical period in which mechanisms mediating the effects of stress on cancer are being pursued in 1976 with the publication of an article in the British Journal of Cancer showing that administration of beta-adrenergic agonists or restraint stress lowers the resistance of growth in the lungs to of systemically-injected tumor cells in rats [80]. Mechanisms have become an important type of causal explanation in the life sciences. Minimally, “[a] mechanism for a phenomenon consists of entities (or parts) whose activities and interactions are organized so as to be responsible for the phenomenon” [81]. A typical heuristic strategy to discover mechanisms is that of decomposition and localization [82]. In this strategy, the causal properties of the composing parts of mechanisms and their organization are supposed to causally alter the properties of a system of interest [83]. It has been argued that this strategy has been widely adopted in the life sciences, such that an acceptable causal explanation of a phenomenon of interest invokes the description of a mechanism, in which the parts’ or entities’ activities act as causes [84].

While the life sciences seem to historically have favored mechanistic causal explanations [8], the mental aspects of stress have been complicated to fit into this framework. One way to conceive “mental mechanisms” is to propose that one replaces the notion of material transformation associated with mechanisms in biology with that of information processing [85]. Thus, “[a] central insight of the mechanistic account of explanation [would be] that biological and psychological phenomena result from mechanisms in which component parts and operations do not individually exhibit the phenomena of interest but function together in an orchestrated and sometimes in a complex dynamical manner to generate it” [86]. While such a perspective can be interpreted to pave the way for interdisciplinary research between psychology and neuroscience, it is important to realize that different appreciations exist of what a complete mechanism would be. Some have favored a view in which higher level mechanisms must be realized by lower level mechanisms until they “bottom-out” at the level of the fundamental laws of physics [87]. Others have instead proposed that “bottoming-out” of mechanisms is determined by the interests and standards of a scientific field or discipline [84]. But, in any case, this puts interdisciplinary research at the risk of never being considered mechanistic enough [88, 89].

Another important aspect of the causal account invoking mechanisms is that it is compatible with what has been called an interventionist “difference-making” account of causation and that has been inspired by the practice of experimental science. According to this account, a candidate cause can become part of a causal explanation of a phenomenon of interest if a so-called ideal intervention on this candidate cause would result in a change of the phenomenon of interest, provided certain background conditions are met [90, 91]. As such, this interventionist account of experimental science contributes to the proposal of mechanistic causal explanations in the life sciences [92, 93]. While it is true that “ideal interventions” indicate causal dependencies but inform little about how causal entities are organized in a mechanism, it is also important to warn against the temptation to merely describe mechanisms without interventions [91, 94]. Therefore, we have proposed that causal mechanistic explanations in (psycho)neuroimmunology should typically require intervention strategies [88]. However, it is also essential to keep in mind that an experimental interventionist strategy on its own is an excellent way to establish “difference-making” or “dependency accounts of causation of the kind ‘A causes B.’” This, thus, illustrates already one way in which the arrangements of the different tiles of causal accounts can contribute to the emergence of a more coherent picture.

If elucidating mechanisms can be considered to provide answers to the questions of how psychological factors contribute to cancer and, conversely, how cancer may affect patients’ mental life, it does not, as such, identify what causal contributions may play a role in an attempt to answer the question of what causes. However, experimental intervention studies can determine factors that make a causal difference for an outcome of interest and can, therefore, be considered to be complementary to mechanistic accounts [88] and enable to propose potential causal factors.

In the early 20th century, epidemiological approaches started to indicate that chronic diseases, such as cancer and cardiovascular disease, were resulting in higher mortality than diseases caused by microbes [95]. In addition, some epidemiologists emphasized that one should not only focus on microbes as a causal agent, but also consider the host and its environment to explain diseases and mortality [96]. Moreover, given the decline of morbidity and mortality due to infectious diseases, they proposed to expand the scope of epidemiology to all diseases affecting populations.

When it comes to the relationship between psychological factors and cancer, one of the first studies proposing an epidemiological approach compared a cohort of over 2,500 individuals with characterized personalities in terms of cancer incidence at a 10-year interval and reported that: “Among women there is a much higher proportion of cancer cases among the substable (personalities) than among the controls (p < 0.005)” [39]. A similarly designed study compared reactions to stress and family attitudes of a sample of about a thousand male medical students between 1948 and 1964 to later development of diseases and illnesses (hypertension, coronary heart disease, cancer, and mental illness) between the ages of 35 and 55 [97]. Interestingly, the incidence of low closeness to parents reported by students was found to be twice as high in men who later developed cancer as compared to the total male population [97].

These and other findings as well as their interpretations have been discussed in a review article published in the first volume of the Journal of Behavioral Medicine making important methodological points. While this article recommends prospective studies to address the links between psychological factors or stress, it also expressed concern about the fact that many studies attempting to relate psychological factors or stress to cancer did not take into account biological or confounding factors or used varying measures of interest [98]. Acknowledging some of these points of criticism, a subsequent study refined the closeness to parents scale and found, in an updated cohort of close to a thousand men, that father-son relationships mainly determined this scale and that the association between low closeness to parents and later cancer development existed even after statistically controlling for possible mediating factors such as tobacco and alcohol consumption [99].

One potential limitation of these initial studies is that they often concern all types of cancer [98]. While this is understandable for the sample sizes used [99], there is also a risk that the contribution of psychological factors to cancer gets diluted if they play a more important role in certain types of cancer. Several trends can be observed in recent epidemiological studies on the links between psychological factors or stress and cancer to address this point. A first strategy consists of comparing the risk of developing a particular cancer type in a group of individuals with a specified mental disorder, assuming that this is related to certain psychological factors or stress, to the expected risk in the overall population. Thus, according to a Danish study, no support for increased incidence of breast cancer was reported in women who had been previously admitted to a psychiatric clinic with affective or neurotic disorders [100].

A second strategy to address links between psychological factors or stress and cancer is to perform systematic reviews or meta-analyses of individual studies on the matter of interest [101]. A landmark work in this respect found that prospective studies indicated that so-called stress-related personality or coping style and negative emotional responses or poor quality of life were associated with higher overall cancer incidence [102]. Regarding specific types of cancer, it was reported that, broadly taken, stress-related psychosocial factors were associated with an increased incidence of lung cancer [102]. Although the effects of stress-related psychosocial factors on overall cancer incidence was reported to persist after controlling for smoking and other high-risk behaviors, it is not known if this also specifically applied to lung cancer. However, systemic reviews and meta-analyses are likely to be influenced by publication bias and fraudulent studies [103, 104]. While it is still too early to a have systematic review or meta-analysis on the question of the effects of psychological stress on the outcome of new treatment options of cancer, such as immunotherapy, it may be worthwhile relating some individual studies. Thus, a recently published clinical trial indicated that emotional distress prior to immune checkpoint inhibitor therapy of treatment of melanoma is associated with reduced 2-year relapse-free survival [105].

A final strategy to study links between psychological factors or stress and cancer is to use large, for example, nationwide databases when they are available, to investigate the relationship between major stressful events (such as childhood adversity, divorce/widowhood, low social integration, or posttraumatic stress disorder) and subsequent cancer incidence. Thus, low social integration and divorce or widowhood were found to be associated with an increased risk of developing ovarian cancer [106]. Other work suggested that the modest increase in cancer incidence after a first-onset stress-related disorder in adulthood is mediated by familial factors as the risk is similar in siblings [107]. High childhood adversity in women was found to be associated with a higher risk of developing breast and cervical cancer in adulthood [108], but it is important to mention that this study did not include a comparison to siblings.

Given that the Journal of Behavioral Medicine initially explicitly invited epidemiologist to submit their work and that this journal has published important articles on the question [98, 99], we propose to start the period during which epidemiological studies addressed the links between psychological factors or stress and cancer in 1978, with the publication of the first issue of that journal. Epidemiological studies can be considered to provide probabilistic notions of causation in the sense that they propose “factors that make a difference to the distribution of disease or to individual health status” [109]. The choice of such factors is important but tends to be those that are deemed of interest by the discipline of the lead author of the study. There are thus risks that the contributions of or interaction with factors other than those of interest to a particular discipline are not considered [98]. Epidemiological studies done by more interdisciplinary teams could mitigate this risk and help identify factors that are closer to making a more important difference for health outcomes. In addition, it is important to distinguish between observation and intervention studies when considering “difference-making” or “dependency” notions of causation. In the former, one establishes a relationship between a factor and an outcome of interest based on, for example, covariation in a population, while in the latter this relationship is obtained after an intervention on the factor of interest. Thus, a population-based survey analyzing correlations between lifestyle factors and health outcomes would be an observational study and a randomized clinical trial (RCT) addressing the effect of a lifestyle intervention on a health-related outcome in a sample of a population would be an intervention study [110‒112].

Here, we have drawn on the notion of mosaic causation to describe causal links between psychological factors and cancer during two different time periods, namely between 1954 and 1976/8 and from 1976/8 to 2023. In the post-WWII era up to the mid-1970s, this question was dominated by psychosomatic medicine and a form of “agency causation” (see Fig. 1). From the end of the 1970s onward, questions regarding psychological factors and cancer were increasingly investigated from a basic science and epidemiological perspective mobilizing mechanistic and probabilistic notions of causation, respectively (see Figs. 1, 2).

It is important to keep in mind that over this period spanning more than 70 years, interest in the direction of the causal claims between psychological factors and cancer has changed from an initial emphasis on psychological factors causally contributing to cancer to approaches investigating how cancer can alter psychological factors. In addition, the consideration of psychological factors has evolved over this period as well, from being heavily inspired by psychoanalysis immediately post-WWII to a more biologically grounded one with the notion of stress and the biologization of disorders like depression and anxiety. Finally, the diagnostic means, etiological theories and treatment options of cancer have obviously changed over the same period, but cancer is still often diagnosed with tissue biopsies, described as uncontrolled growth and continues to be considered a potentially treatment-resistant fatal disease.

Regarding the different types of causality that were introduced previously, it is important to remind ourselves that all research on the links between psychological factors or stress and cancer ought in the end to concern individual patients. This is perhaps most clearly put in practice by initial psychosomatic medicine that proposed individual “agency causation” and grouping of individual characteristics. Instead, mechanistic and epidemiological research studies are based on samples and statistical approaches. Nevertheless, they are both considered to be relevant for individual patients, even though it is rarely made explicit how this would be the case. In addition, research into mechanisms linking psychological factors or stress and cancer has often been done in animals, thus raising additional questions about translational relevance.

Another important distinction when it comes to types of causality in the life and health sciences is between difference-making causal theories and mechanistic accounts. While initial psychosomatic medicine put forward the idea that psychological factors make a difference in the development of somatic disease, it did, at best, allude to mediating mechanisms. However, it did not seem able to propose ways to operationalize and test if and how such mechanisms would actually play a role in mediating the effects of psychological factors on cancer. The notion of stress allowed to propose some mechanisms, but at the price of giving up the specificity of psychological factors dear to initial psychosomatic medicine. As of today, a consensus has emerged to say that there is no “cancer-prone” personality, and that stress and life events are not to be considered major causes of cancer [113]. Nevertheless, it is important to recognize that social isolation and deprivation along with depression can negatively influence the quality of life and survival after a cancer diagnosis [113].

A related, but distinct, point of interest that deserves to be discussed here is that between the causation brought by a whole (molar causation) and that produced by parts of a whole (molecular causation) [24]. This discussion resonates with that of up- and downward causation [114, 115]. Upward causation is familiar to most actors in the natural sciences as it proposes to causally explain a phenomenon of interest at a perceived higher level of organization by appealing to activities of entities at lower levels. This would correspond to the ways that cytokines or cells reflecting tumor growth act on the brain to alter psychological functioning but also to the action of certain hormones, released as a result of stress, on tumor growth. Conversely, according to supporters of downward causation, entities at perceived higher levels of organization display new causal powers compared to entities at lower levels that affect their constituting components, “so that lower level events take place differently within them” [114]. Some authors have thus proposed that downward causation can be viewed as a partial constraint on lower level entities, for example in mechanisms [114‒116]. This would be one way, which could include or go beyond the action of specific hormones, such as corticosteroids and catecholamines, to envision the effects of psychological factors or stress on cancer.

In hindsight, psychosomatic medicine can be considered to have set the scene in terms of formulating and framing questions that were provocative regarding the “agency causation” of an individual’s “psychological set-up” contributing to cancer. These questions have then been operationalized in complementary ways by basic science, on the one hand, and epidemiology on the other hand, mobilizing mechanistic and probabilistic notions of causation approaches, respectively. For example, it was found that many studies that attempt to provide further mechanistic detail about the links between psychological factors or stress and cancer cited epidemiological evidence in their introductions. Indeed, epidemiology can identify potential causes, while basic science can determine how these causes work. However, it is important to realize that a statistical account of causation often depends on the list of factors that can be identified beforehand while mechanistic accounts do often not consider the context in which mechanisms operate. If the evolution of causal accounts and their interaction when it comes to the link between psychological factors and cancer time may just have happened to have played out this way historically, it could also reflect the way in which, we would like to see other questions being approached.

In this context, it is important to keep in mind some of the assumptions associated with different causal accounts. Mechanisms, for example, often follow the heuristic strategy of decomposition to propose some component’s activities as causes [82, 83]. While this strategy can easily be applied to non-living aggregate systems, this is not necessarily the case for integrated living systems characterized by many feedback loops [117‒119].

Interestingly, observational studies can play a heuristic role in identifying possible mechanism components when transposing a mechanism schema to new fields [120, 121]. Epidemiological studies are typically observational and investigate the relationships between categories of putative causal factors and a health outcome of interest in a population with a long-standing consensus that such relationships are not causal, but correlational [122, 123]. Recently, a number of epidemiologists have taken specificity of association in observational studies to indicate causal relationships [124, 125]. However, emphasizing specificity comes with a risk that studies that are closer to experimental studies, in the sense of better corresponding to well-defined potential interventions, are valued over observational ones that are less specifiable in this way, even though observational studies are, by definition, non-interventionist [126, 127]. Alternatively, it has been argued that the statistical testing of formulated causal hypotheses regarding mechanisms, which can account for the correlations observed and indicate potential causal mechanisms or mechanism schemas [128]. Thus, mechanistic and observational studies can play complementary roles in determining causal relationships.

It is interesting to compare the proposal that causal claims in the health sciences ought to be based on evidence of both mechanisms and probabilistic dependencies [25] and the more freely formulated idea that “the full array of causal concepts developed within philosophy could very well be of interest to practicing scientists” [24]. So, while it is probably always good to go beyond the search for and the commitment to “the-one and only” theory of causality, typical of many academic disciplines, it remains an outstanding question as to whether a particular articulation of causal accounts is preferred over another. This naturally applies to other chronic disease conditions than cancer as well. While chronic inflammatory or neoplastic conditions are often thought to be of multicausal origin, we would like to propose here to also consider multiple causal accounts in our explanations of such conditions. Thus, the interplay between multicausal etiology and causal mosaic can occur anywhere between one causal account suggesting multiple causal factors and multiple causal accounts indicating one causal factor to many causal accounts suggesting many causal factors.

While we agree with the idea of taking into account different causal approaches, we would like to go further and also consider how the different accounts of causation are related and can be articulated. In addition, it is important to keep in mind that, when it comes to research, causal claims are frequently context- or model-dependent and that empirical evidence comes in degrees [24]. Thus, one can promote pluralism by requiring that different methods are used to approach causation. However, these different accounts often respond to different questions regarding causality between what is causing and how it is causing. Accordingly, a mosaic view only makes sense when it gives rise to images giving new insights by arranging the tiles in certain ways.

Different causal explanations, like epidemiological evidence and mechanisms, reinforce each other, and should not be taken as a criterion for exclusion. For example, while some social factors, related to socio-economic status, as indicated by epidemiological studies, can be shown to be biologically embedded or have biological correlates [129], not all social factors can easily be transposed into biological consequences. However, there is no reason to exclude the latter simply because the “grain” of analysis would not correspond. Indeed, the social factors indicated by epidemiological evidence that are not (yet) biologically embedded may well be fruitfully articulated with social mechanisms.

Mid-20th century psychosomatic medicine can be considered to have put the “agency causation” idea that a person’s psychological “set-up” can contribute to cancer on the agenda of some scientific domains. Since the 1970s, after the formulation and operationalization of stress responses, mechanisms involving activation of the sympathetic nervous system and the hypothalamo-pituitary adrenal axis, resulting in catecholamine and corticosteroid hormone release, respectively, have been proposed, in progressively more molecular detail, to explain the effects of psychological factors on tumor growth. Conversely, more recently, mechanisms involving cytokines have been hypothesized to explain how cancer and its treatments can affect mental well-being. Epidemiological work from the end of the 1970s onward indicated that stress-related psychological factors are associated with higher overall cancer incidence, including lung cancer [102], thus suggesting that psychological factors can contribute to the course of cancer. Retrospectively, this historical order of events between initial “agency causation” followed by mechanistic and probabilistic notions of causation being developed in parallel, seems to have worked well as a causal mosaic for the question of the links between psychological factors providing progressively more nuanced responses. However, this does not mean that different causal tiles should always be arranged in this way to gain insight into questions. Instead, the arrangement of these tiles should be adapted to the (explanatory) needs of the research field.

The authors would like to sincerely thank Federico Boem (University of Twente), Maarten Limper (University Medical Center Utrecht), Abigail Nieves Delgado (Utrecht University), and Federica Russo (Utrecht University) for their constructive feedback and the organization of a workshop based on an earlier version of this paper.

The authors have no conflicts of interest to declare.

This study was not supported by any sponsor or funder.

I.A. has contributed to drafting and editing of the article and has approved its final version. J.P.K. has contributed to the drafting, writing, and editing of the article and has approved its final version.