Abstract
Population screening for early-stage cancer or cancer precursors began in the mid-twentieth century, with the goal of reducing suffering from cancer illness and lengthening average life by preventing cancer deaths. Since the establishment of cancer screening, concerns have emerged that it may be doing considerable harm; despite this, screening practices have remained relatively intractable. This intractability in the face of harm is the central problematic of my analysis. I reinterpret a large study of breast, cervical and prostate cancer screening completed recently by our Australian research group, working across empirical bioethics, public health and social science. I suggest three reasons why cancer screening might persist as it does, and thus reach conclusions about what might be required to make cancer screening systems more responsive to the potential for harm.
Keywords
During half a century, early detection screening imposed indiscriminately on the entire population to detect asymptomatic cancer has been highly promoted, almost glorified, as a strategy to improve cancer control …. Over time, both the scientific and the health care community have been surprisingly unwilling to embrace accumulating evidence that wide population-based early detection screening for cancer has … induced considerable harm to a large population of healthy individuals.
—Adami et al. (2019, p. 4)
The quote above comes from a 2019 Editorial in the European Journal of Clinical Investigation. The authors are unusually direct, arguing strongly that screening for early-stage cancer—especially for breast and prostate cancer—should be abandoned because it is doing harm. They worry that although these harms are well-recognised, the general public, clinicians and policymakers remain enthusiastic about screening. Here I focus on this perplexing problem—the tenacity of cancer screening in the face of strengthening evidence of harm—and explore some possible explanations for its persistence.
Some now-axiomatic observations about screening activities include that they seek to: recruit healthy populations to periodic testing, sort people into those likely versus unlikely to have the condition, and direct those most likely to have the condition to diagnostic testing (Armstrong & Eborall, 2012; Carter, 2016). Cancer screening, like all screening, is thus the first step in a cascade that can turn participants into cancer patients (Carter, 2016). The claimed purpose of cancer screening is to reduce suffering and death from cancer, while avoiding introducing additional harm, at a reasonable cost (see Australian Government: Department of Health [n. d.]). Benefits rely on preventive or curative treatment occurring earlier in the disease course, making intervention easier to tolerate, less harmful and more effective.
How Did Cancer Screening Practices Develop?
The earliest example of cancer-related screening—well known to health sociologists—is Pap testing for the precursors of cervical cancer (Armstrong & Eborall, 2012). This began in the 1940s (Carter et al., 2015b), and gained traction through the 1960s and 1970s (Williams, 2016). As shown by Casper and Clarke (1998) in their now-classic analysis of ‘Making the pap smear the right tool for the job’, the establishment of cervical screening was highly contingent. It also created cultural expectations that cancer was a disease that could be addressed via screening: it arguably made screening the right tool for the job of the war on cancer.
Pap testing was followed by mammographic screening for early breast cancer, which was ad hoc from the mid-twentieth century and organised from the 1980s onwards (Carter et al., 2015b). Next came the Prostate Specific Antigen (PSA) test, first approved for prostate cancer risk screening by the US Food and Drug Administration in 1994. Without formal ‘organisation’, PSA testing became a de-facto population screening programme in many jurisdictions (Carter et al., 2015b). In the twenty-first century, bowel cancer screening has been introduced in many high-income countries (Flitcroft et al., 2010). The technologies for cancer screening continually shift, but across all four of these conditions—cervical, breast, prostate and bowel cancer—formal or informal population screening remains in place as a ‘right tool for the job’ of reducing cancer mortality, making them an important case for understanding socio-technical screening systems. 1
Cancer screening is a complex, contingent and situated activity. It is historically emblematic (Armstrong, 1995), resource intensive, morally compelling for participants (Armstrong & Eborall, 2012) and technocratic. I am writing from the interdisciplinary hinterland between the social sciences (e.g., health sociology, science and technology studies), health ethics, medical history, public health and epidemiology. In my view the social science literature has not come to grips with the epistemic complexity of screening or paid close attention to the problem of harm; conversely the public health literature has produced relatively thin and atheoretical discussions of the structural and cultural dimensions of screening, and has tended to assume that more and better evidence will solve the problem. By working across these literatures, I aim to join up different ways of seeing cancer screening and to provide new insights into the inability of both health systems and citizens to respond to the harms done by screening.
This article proceeds in four steps. First, I outline previous research, the basis for this article. Next, I outline the central problematic: harm caused by screening. Third, I introduce what we already know from the public health and social science literatures about cancer screening, especially regarding harm, and lay out three additional, complementary explanations for the persistence of screening. Finally, I draw some conclusions about what might be required to take harm seriously in cancer screening.
Our Research on Cancer Screening
This article draws heavily on, and re-theorises, prior empirical and theoretical research on cancer screening completed at the Centre for Values, Ethics and the Law in Medicine (now Sydney Health Ethics) at the University of Sydney from 2012 to 2017. This included broad-based research, including our analysis of tensions in the evidence base (Carter et al., 2015b) and our definitions of overdiagnosis (Carter et al., 2015a; Carter et al., 2016a), but focused on an empirical examination of three cases—breast screening, cervical screening and PSA testing—especially in Australia.
Our work on breast screening, led by Dr Lisa Parker, asked how Australian programmes and guidelines have been formed; 33 clinicians, programme managers, policymakers, advocates and researchers involved in these processes were interviewed in-depth, with analysis strongly informed by ethics theory (Parker et al., 2015). Our work on Australian cervical screening, led by Dr Jane Williams (2016), first systematically examined: a) the information provided to Australian women, and b) the construction of ‘organised’ screening in the peer-reviewed literature. Williams then collected data from 22 policy actors who had participated in cervical screening guideline-setting processes in Australia and internationally, and gathered public-record reports of Australian guideline-setting. Analysis drew on ethical and epistemic concepts. Our work on PSA testing was led by Dr Kristen Pickles (Pickles et al., 2016), who interviewed 40 general practitioners (GP) in Australia and 29 in the UK about their practices and reasoning, guided by grounded theory methodology. This illuminated GPs’ varied logics of care, the values underpinning these, and important structural and cultural differences between the two jurisdictions.
Finally, we have run several community juries on cancer screening, led by Dr Chris Degeling (Degeling et al., 2018). These processes, founded in theories of deliberative democracy, bring inclusive groups of 10–20 people together over two days, present them with evidence and argumentation, and support them to debate and make recommendations about a contentious policy problem. We have conducted juries on PSA testing and breast screening in older women. More methodological information about specific sub-studies can be found in the primary papers.
The Central Problematic: Harm Caused by Screening
Overdiagnosis occurs when a diagnosis is being used in a population in a way that would be considered correct by relevant professionals, but this produces an unfavourable balance of benefits to harm (Carter et al., 2016a). In this context, overdiagnosis occurs when, as a result of screening, cancer is being diagnosed correctly, and this is causing sufficient harm to be of concern. Overdiagnosis occurs in cancer screening largely by detecting cancers that did not need to be found because they would never have progressed, or would even have regressed, if left undetected. 2 For such cancers, all treatment is unnecessary treatment; arguably, the greater the harm done by this treatment, the greater the ethical problem it poses (Carter & Barratt, 2017).
Estimates of the extent of overdiagnosis in cancer screening are sobering. For screening mammography, a best-case estimate suggests that over twenty years of screening, for every two women prevented from dying early from breast cancer, one is overdiagnosed; a worst-case estimate is that for every one woman who avoids breast cancer death, ten are unnecessarily diagnosed and treated (Brennan & Houssami, 2016; Gøtzsche & Jorgensen, 2011; Marmot et al., 2013). With respect to prostate cancer, on one estimate, over thirteen years of annual testing, approximately twenty-seven men are diagnosed unnecessarily for each prostate cancer death avoided (Barry, 2018); a recent Australian study estimates that 41 per cent of prostate cancer is overdiagnosed (Pathirana et al., 2019). While cervical cancer itself is not thought to be overdiagnosed to any degree, many women treated for cancer precursors would never have developed cervical cancer, and these treatments can harm (e.g., making it more difficult to carry a pregnancy to term) (Chan et al., 2012). People overdiagnosed as a result of cancer screening are not only injured by being recruited into unnecessary treatment: they also experience, for example, opportunity costs and psychological harms, which are rarely measured (Harris et al., 2014).
In addition to the harms of overdiagnosis, screening programmes unavoidably produce some false positives and false negatives (Armstrong & Eborall, 2012). The rate of these depends on factors including breast biology, programme design and the screening technology used (Carter & Barratt, 2017). Although screening services tend to tolerate false positives (more than false negatives), false positives cause injury (e.g., from biopsy) and significant psychological and existential harms (Brodersen & Siersma, 2013). As David Armstrong (2012) has noted, the potential for ‘negative psychological reactions’ to screening—including to false positives—has been recognised since the 1980s.
In summary: cancer screening actively recruits healthy people, benefits some, and causes harm to perhaps many others. Despite this, when scholars such as Adami et al. (2019) suggest that screening should be withdrawn, they are often accused, to quote Quanstrum and Hayward (2010), of having ‘a callous disregard for…life and well-being’ (p. 1076). It is this tension that I turn to now.
Explaining The Persistence of Cancer Screening Despite Its Potential to Harm
My goal is to draw out some possible explanations for the intractability of screening, situating this in two literatures—the overdiagnosis literature from public health and healthcare and the literature on the sociology of screening.
In 2017, Pathirana et al. developed a descriptive synthesis of drivers of and solutions for overdiagnosis. In 2019, Armstrong synthesised the social science literature on screening. While neither claim to be comprehensive, in my view they provide ready summaries of two key, but often disconnected, literatures relevant to my central question. From public health and healthcare services, Pathirana et al. asked why correct but harmful diagnosis was occurring and what could be done about it. As much overdiagnosis occurs as a result of screening or opportunistic testing, ‘overdiagnosis’ and ‘harm from screening’ overlap significantly. The authors propose that overdiagnosis occurs due to drivers in five domains: culture, health system, industry and technology, professionals, and patients and public. These seem to me to reduce to: ‘human frailties’ (e.g., cognitive biases, false beliefs and emotional responses), ‘technological cause’ (the increasing sensitivity of diagnostic tests), ‘practice challenge’ (the complexity of care) and a cluster of ‘structural causes’ (e.g., processes for practice guideline formation and the financial arrangements of healthcare). This synthesis is descriptive and broad rather than deep, but is helpful in that it seeks to capture as many relevant factors as possible.
Armstrong (2019) asked what social theory-informed analysts have contributed to the sociology of screening. She notes several broad trends. Sociological work on power and the responsibilisation of health suggests that screening participation can be experienced as a moral and social obligation. People rarely participate in screening with full understanding of the potential implications. Social scientists have focused on the experience of participation more than on professional practices and perspectives. Armstrong also highlights contributions of key scholars. For example, Armstrong (1995, drawing on Acherknech) mapped historical transitions towards ‘surveillance medicine’ from the twentieth century, characterised by monitoring of whole normal populations, blurring of the distinctions between health and illness, and a focus on chains of extracorporeal ‘risk factors’ rather than on underlying disease. Armstrong (2012) theorised screening as a form of surveillance which ‘opens up a space of possibility’, a temporal dimension that did not characterise former epochs in healthcare (pp. 401–403), and developed this temporal understanding of screening in later work. Later, Gillespie (2015) coined the concept of ‘measured vulnerability’, explaining how the statistical risk values generated through PSA testing induced perceptions of impending disease and uncertainty, increased medical contact, restructured everyday routines, and altered social relationships.
The sociological literature offers considerable insight for scholars seeking to understand screening participation and practices. Notably however, Armstrong (2019) relies heavily on the public health literature to talk about overdiagnosis and harm, suggesting a relative dearth of social science research on these questions. The social science literature tends not to ask why screening continues, seeking instead to explain how it works. Conversely, the public health literature has proposed a range of possible ‘causes’ of overdiagnosing practices which are pragmatically and descriptively useful, but are relatively under-theorised. On this background, I connect our primary research findings with the public health, social science and applied ethics literatures to provide new answers to the problem of the persistence of screening in the face of harm. I will propose three central problems, which I have called ‘the agency in systems problem’, ‘the epistemic problem’, and (after Beckert, 2016), ‘the problem of fictional expectations’. A critical consideration of these cultural and social dimensions of cancer screening, I will argue, may help explain the intransigence of screening.
The Agency in Systems Problem
Each form of cancer screening relies on a complex and relatively contained socio-technical system. The point of care imbricates different test technologies (capturing images, blood or cells), frontline clinicians (e.g., technicians in mammography or GPs in PSA and cervical screening), reading and interpretation systems (including technicians, pathologists, radiologists and information technology), and communication technologies. Onto this system of direct service provision are attached dependent and synergistic systems for data collection, evaluation, funding, guideline formation, specialist treatment networks (e.g., breast physicians, urologists, gynaecologists and oncologists), monitoring and support systems following treatment, and vast structures of fund-and-awareness-raising that are particularly prevalent in relation to cancer, driven by both non-profit organisations and government. 3
Material and human agency each contribute to such socio-technical systems. Here I am following Leonardi’s (2012) conception of socio-technical systems, as containing both social and technical subsystems, which mutually shape one another. The social subsystem includes elements such as roles, power relations and status systems. The technical subsystem contains both persons and artefacts, and it brings the intentionality of the persons and the materiality of the artefacts together, a meeting and co-influencing of human and material agency. Leonardi takes social agency to be ‘coordinated human intentionality formed in partial response to perceptions of a technology’s material agency’, and the material agency as ‘ways in which technology’s materiality acts…[which is] activated as humans approach technology with particular intentions’ (Leonardi, 2012, p. 42). I am conceptualising intention more broadly than simple intention to perform specific actions: human intention includes goal directedness, values and evaluation, and shared purpose. This brings a set of interconnected findings from our cancer screening study into play.
We found that implicated actors within socio-technical screening systems (e.g., those who made guidelines, did research and provided frontline services) had strikingly divergent goals and values in relation to cancer screening. Williams et al. (2017), for example, showed divergent conceptions of what cervical screening was for, ranging from finding and eliminating every possible case of cervical cancer, to ensuring access to screening, to preventing excessive screening so as to avoid unnecessary harm. Parker et al. (2015a) showed widely differing understandings of concepts such as avoiding harm and delivering benefit among breast screening experts; and Pickles et al. (2015) showed very different priorities among Australian GPs about PSA testing, ranging from minimising harms from underdiagnosis to minimising harms from overdiagnosis. These differences have dramatic implications for screening organisation. While the public face of screening programs is unified, emphasising the rational application of evidence to reduce suffering and save lives, the human intentionality within these sociotechnical systems is in fact fractured and sometimes conflicting.
We also found that the macro/meso structuring of screening systems profoundly altered the limits within which human agency was expressed. Pickles et al.’s (2016) study identified very different PSA testing regimes in different jurisdictions. The UK featured centralised (National Health Service) guidance against screening asymptomatic men, less demand for testing, delay routines (men who asked for testing were given information only; testing was deferred to a second appointment) and well-defined referral pathways. In contrast, at the time of our study 4 Australia featured very different advice from different actors (e.g., local urologists, government departments, cancer charities and medical colleges), a fee-for-service funding model that financially rewarded testing, and GP beliefs that they were at medicolegal risk. UK GPs were bemused at the idea of actively offering PSA tests to asymptomatic older men. For Australian GPs, actively offering the test was normalised, resisting testing was personally burdensome and worrying, PSA testing decisions were part of everyday work, and clinicians had generated widely divergent personal approaches to testing (Pickles et al., 2015, 2016).
Organised Australian screening systems for cervical and breast cancer, meanwhile, pressed actively towards maximising participation. Screening programmes are evaluated against aspirational participation rates: more participation is better (Australian Government: Department of Health and Ageing, 2009). Screening communication campaigns tend to emphasise or inflate benefits and minimise harms, for example, Williams et al. (2014) found that cervical screening communications neglected harms and provided misleading benefit information (e.g., using relative risk estimates, which inflate perceptions of screening benefits). Parker et al. (2015b) found a majority of expert respondents thought women should be actively guided towards screening; about half of these thought women should not be informed about overdiagnosis, a central rationale being participation maximisation. The main message from screening services is ‘you should participate’; unsurprisingly, people have absorbed these messages. Australian (Moynihan et al., 2015) and American (Schwartz et al., 2004) survey research has shown that people are generally enthusiastic about screening, believing participation to be both beneficial and in some cases a moral obligation, consistent with social science insights about responsibilisation, and the roles, power relations and status systems in the social subsystem of screening services (Armstrong, 2019). In addition to normalisation and feeling obliged, people also attend screening to seek reassurance (Whynes et al., 2007).
These observations can be brought together in an explanation. System structuring offers considerable resistance which can encourage or discourage testing. UK PSA testing was a divergent case in our study: centralisation and provision of resources and clear care pathways focused human agency in the system, and significantly shaped testing culture. Australian screening systems (organised or otherwise) pressed towards maximising participation. Notably, even in explicitly designed systems, human intentionality was widely divergent, and it was more difficult to hold critical than uncritical views of screening. In these contexts, the material agency of the technology—mammogram, pap smear, PSA test—tended to dominate: established systems pressed towards the technology being applied to as many people as possible, consistent with Armstrong’s (1995) conception of surveillance medicine.
The Epistemic Problem: The Impossible Necessity of Evidence
The public health ideal is that evidence should distinguish between justifiable and unjustifiable screening activity. On this view, experts need to only agree on the proper methods for synthesising evidence, and this will ensure trustworthy, high quality ‘evidence based’ screening (Williams et al., 2020). Unfortunately, cancer screening has an ‘epistemic problem’. Evidence is treated as necessary, but is both contested and inaccessible. This can be thought of as ‘the impossible necessity of evidence’: it seems particularly problematic for cancer screening. 5
Contestation
The evidence about cancer screening is strongly, publicly contested. Despite the efforts of dedicated epidemiologists, it is impossible to answer the question—‘does this kind of screening do more harm than good?’—in a way that all or even most stakeholders would accept (Carter et al., 2015b). Methodological experts disagree over: how to measure outcomes, which studies should be included or excluded in syntheses of the literature, and on how to work with systematically biased data. 6 Researchers’ prior commitments to screening have been shown to correlate with their estimation of the extent of overdiagnosis (Hofmann, 2018), and methodological disagreements provide ample space within which policy actors—such as departments of health, specialist clinical colleges, disease advocacy groups, and individual influential clinicians—can provide conflicting advice. This creates great uncertainty, which as noted by Armstrong (2019), is a central feature of the screening landscape.
Han et al. (2011) have usefully distinguished between three sources of uncertainty in healthcare. Probability uncertainty arises from the nature of risk-based or probabilistic information. Complexity uncertainty arises from the multifactorial nature of cause in individual cases. Ambiguity uncertainty arises from imprecise, conflicting or absent information. I have already mentioned our comparative study of PSA testing in the UK and Australia (Pickles et al., 2016). UK GPs, working in a structured system, did not feel particularly burdened by uncertainty. In Australia, where experts’ advice conflicted, GPs explicitly discussed strategies for managing ‘ambiguity’ uncertainty. Some concluded that science was no help, developed their own intuitive protocol and shielded patients from uncertainty. Some delegated decision-making to others. Others worked through the uncertainty with colleagues and patients. The latter looks most like the shared decision-making often advocated as a strategy for dealing with uncertainty (Han et al., 2011), but clinicians taking this route often worried that they were overburdening patients. GPs expressed strong frustration with the ambiguity uncertainty in the system. Similarly, in our community juries with older women about mammography, participants expressed frustration with ambiguity uncertainty, expecting that scientists should be able to measure screening outcomes, and doctors should be able to advise an individual as to whether a screen detected cancer is overdiagnosed (Degeling et al., 2018 ). Unfortunately, this is rarely possible, partly because of epistemic inaccessibility.
Inaccessibility
Knowledge about overdiagnosis of cancer is inaccessible for two reasons: the limits of epidemiological method and the counterfactuality or non-identifiability problem (Carter et al., 2016a; Hofmann, 2014), which I address in turn below.
Epistemic inaccessibility arises in part from the limits of epidemiological method. As noted, cancer screening evaluation tends to focus on a relatively narrow band of outcomes: mortality, false positives and false negatives, and basic information about morbidity. However, the range of possible harms is wide. It includes psychological harm, opportunity costs and, in some jurisdictions, significant financial harm—including medical bankruptcy (Harris et al., 2014). These harms are arguably insufficiently recognised and measured, but addressing this is a potentially mammoth task. In one discussion with experienced screening epidemiology colleagues, we broadly agreed on the diversity and importance of screening harms. I then asked innocently how these harms are, or could be, included in calculations of screening outcomes. My colleagues fell silent, then said the task was too vast, and that if it were possible it would immediately demonstrate that no screening programme was justifiable. Even the best evidence, then, is a considerable oversimplification of the complexity of screening outcomes. Processes of constructing evidence also tend to focus on relatively straightforward questions about the screening test technology (does this test work?) and the best screening interval (how often should people be tested?), and less on the complexities of practice arrangements, which are central to participants’ experience of screening (Williams et al., 2020). The arrangement of screening services varies enormously between jurisdictions and screening technology changes, so what appear to be comparable data may rest on profoundly different conditions and assumptions (Carter et al., 2015b). In short, screening outcomes data are inaccessible in part because of the complexity of the conditions of their construction and the epistemic limits of method.
I now turn to the second kind of epistemic inaccessibility proposed above: the counterfactuality or non-identifiability problem. Recall that people diagnosed with cancer through screening will have been recruited because they were ‘healthy’ members of the population. If screening leads them to be diagnosed and receive cancer treatment, they become a cancer patient. Once diagnosis and treatment occur, it is not possible to know what would have occurred ‘without’ diagnosis and treatment. Thus helpfully diagnosed individuals cannot be distinguished from overdiagnosed individuals, and overdiagnosis cannot be observed and counted in ways that, for example, the number of people diagnosed with a condition can be observed and counted. 7 Overdiagnosis can thus generally only be estimated in populations (rather than directly observed or experienced), and this has real flow-on effect in people’s reasoning. As mentioned, Parker et al. (2015b) found that most breast screening policymakers thought women should not be told about overdiagnosis harm. What reasons did they give to justify this non-disclosure? They argued: overdiagnosis was ‘not a real harm’, women don’t ‘experience’ overdiagnosis as a harm and population-based information was ‘not relevant’ to individuals. Essentially, they argued that harms you cannot directly observe or experience are ‘not real harms’, so should not be taken seriously.
In this section, I have proposed that the harms of screening are not taken seriously, so screening persists, in part because of the ‘impossible necessity of evidence’. To support this contention I have shown that evidence is supposed to adjudicate between worthwhile and problematic screening, but methods for evaluating screening evidence are deeply contested, frustrating clinicians and citizens; epidemiological methods cannot incorporate the range of harms caused by screening; and because cancer overdiagnosis can rarely be directly observed, decision-makers can dismiss it as not ‘real’ or important. Thus, the impossible necessity of evidence supports the persistence of screening despite its potential to harm.
The Problem of Fictional Expectations: Screening is Aligned to the Cultures of Late Modern Capitalism
Finally, using Beckert (2016), Petersen and Krisjansen (2015) and Rajan (2006), I will suggest that cancer screening persists because of its broader cultural resonance. Beckert’s (2016) economic sociology emphasises the importance of imagined futures in late modern capitalism. Capitalism, he argues, is built on a belief that we face an open and uncertain future; this belief is an under-recognised causal force. People mostly do not make decisions by weighing up determinant outcomes or even known probabilities: decisions are usually made under conditions of profound uncertainty. Under uncertainty, we need something to guide action: Beckert proposes that ‘fictional expectations’ provide this guidance, causing people to act in the present by pulling them into alternate futures.
‘Fictional expectations’ sit somewhere between fact and imagination, weaving the two together in a way that makes them more trustworthy and compelling. They can be shared and so can coordinate action. They are expressed through compelling narratives, but are also serious (unlike completely made-up stories). Like authors studying bio-capital (Petersen & Krisjansen, 2015; Rajan, 2006), Beckert (2016) emphasises the promissory quality and hype entailed in fictional expectations: they emphasise potential benefits rather than harm. Fictional expectations are also political: contests over which expectation of the future will dominate are contests over how resources should be allocated. Financial-market forecasts, for example, or pitches to venture capitalists pivot on fictional expectations. In this sense, like promissory discourse generally, they are performative (Petersen & Krisjansen, 2015). 8 ‘Fictional expectations’ shape human perceptions of how the world works and impart transcendental qualities to goods and services. Consumption relies on fictional value in Beckert’s account in ways well-rehearsed in social theory, as he acknowledges. 9 Transcendental promises made to consumers are imaginings about how different the future could be. Goods and services are consumed less for their material qualities and more for their symbolic connotations, and the way they are expected to position the self in relation to others. These promises, also, are fictional expectations, competing for consumer attention.
The intractability of cancer screening seems to arise in part from its resonance with late modern capitalism. In the most obvious sense, in many jurisdictions healthcare is a competitive market. In Australia, for example, Collyer et al. (2015) have shown the steady restructuring of healthcare towards private and for-profit service provision from the 1980s, particularly in screening-related specialities such as radiology and pathology. The private sector dominates production and sale of screening technologies, and the ‘screening specialities’ are, to quote the president of the Radiological Society of North America, ‘fierce early adopters of new technology’ (Keen & Massat, 2018, p. 20). These technologies are expensive, creating pressures to make good on the cost (Baker, 2010). New technologies are sold to service providers as innovative enhancements of technical capability; their uptake by some providers creates competitive pressure on other service providers. 10 These technological fictional expectations are at odds with the idealised model of healthcare which requires rational, rigorous and retrospective evidence before adopting technology.
Earlier I discussed the ‘just get tested’ message that has dominated decades of public health communications, and amplified the threat of cancer risk. Psycho-oncological research has shown that cancer is feared because it is seen as unpredictable, intentionally harmful and hard to contain (Vrinten et al., 2017). This sense of cancer lurking in an open and uncertain biofuture increases the value of being ‘reassured’ that one is cancer-free, which as noted is a primary motivator for participation. 11 But the selling of participation often goes beyond personal reassurance to the kind of symbolic positioning that Beckert (2016) describes. Screening messages, for example, often imply a distinction between the desirable state of having participated and the undesirable state of having not participated (Williams et al., 2014), much as consumer goods and services invite individuals to transform their identities and set themselves apart via consumption. More broadly, cancer charities conjure utopian future expectations: the ‘vision’ of Cancer Council Australia, for example, is ‘A world without cancer’, the kind of compelling fictional expectation—imagination decorated with fact—that might encourage donations and attract government funding and public support (Cancer Council Australia, 2019). 12 Such communications are competitive (for consumers and investors) and so, as Beckert would suggest, they are political. This analysis resonates not only with the temporal dimension of Armstrong’s (1995) surveillance medicine and the governmentality analyses of screening programmes discussed by Armstrong (2019) but also highlights potential concordance between screening and late modern capitalism.
Analysing the persistence of cancer screening using Beckert’s (2016) thinking tools counteracts the false expectations built into the rational evidence-based ideal, and loops us back to the impossible necessity of evidence. After Beckert, and as argued in the previous section, screening decisions cannot be based on known probabilities. Rather, both screening service organisation and participation occur under conditions of profound uncertainty, and also under particular structural conditions; especially the marketisation and consumerisation of health services. Implicated actors—clinicians, patients, citizens and corporations—are living with the promise of an open and uncertain future, and are acting to manage that uncertainty. Interventions are driven by a competition between imaginative visions of the distant future, more than by sober consideration of potential harm in the short term. And when risk reduction is sold to consumers with the promise of future symbolic and positional value—and again, little indication of the deflating uncertainties and trade-offs that might be entailed—it is no wonder that at least some consumers are receptive.
Conclusions
I began with the frustrated call from Adami et al. (2019) for an evidence-based, rational abandonment of some cancer screening. I agree with their conclusions, but think it is unlikely that their hoped-for outcome will happen for rational and evidence-based reasons. I have provided three possible explanations for the intractability of cancer screening. First, screening systems tend to contain pressures towards screening and offer considerable resistance. If human agency and intention are fractured (true in all of our cases except UK PSA testing), system resistance wins out. Second, evidence is necessary, but both contested and inaccessible, which allows screening proponents to actively undermine the ontological security and significance of screening harm. Finally, screening proponents have built convincing, promissory fictional expectations that locate screening as an identity-shaping consumer service and a collective effort to control biofutures.
The epistemic challenges described seem unresolvable. Methodologists cannot agree. Even if they could, epidemiological methods cannot reflect the diversity of screening harm, and because of counterfactuality, overdiagnosis harm can rarely be directly observed. I am committed to the importance of producing the best possible evidence (it was the diligent work of epidemiologists that first uncovered overdiagnosis). But if evidence is a necessary impossibility, understanding the sources of impossibility I have described are fundamental to understand the screening landscape. In this context, evidence can point to a potential problem, but seems unable to offer a reliable solution.
Could the agency in systems or fictional expectations be potential sites for change? Comparing the UK and Australia showed us that systems constrain human agency, nudging away from or towards the material agency of screening technology. Another core finding from our work was that screening systems give the appearance of unified rationality, but contain profoundly fractured human values and intentions. This creates fertile ground for material agency to dominate in large, complex and well-sedimented systems, and puts immense pressures on individuals who push back (e.g., Australian GPs who experienced considerable personal burden when resisting PSA testing). It also shows what coherent and centrally-coordinated policymaking can achieve, and likely points to natural limits of the generalisability of the UK experience.
Beckert’s theorisation complements this explanation: screening persists via persuasive fictional expectations promulgated by services. In a context where everyone is a ‘healthcare consumer’ (rather than a patient or a citizen), one’s response to the challenge of an uncertain and open future is to differentiate oneself by consumption of services that offer future symbolic and positional value; little wonder then that consumers rise to defend these services when they are threatened. Armstrong (1995, 2012) emphasises the temporalisation of disease via screening and other surveillance activities. My analysis contributes a different kind of time horizon: a temporalisation of attention, drawing the gaze away from present, embodied, experienced symptoms, and towards uncertain futures and a promise of self-differentiation; a promise that, in a terrible irony, leads some healthy people to experience unnecessary present harm.
Being more responsive to screening harm would require radical shifts in the focus of health systems: from risk to suffering, from consumerism to care, from expansive market competition to a central concern with possible consequences. Such a shift would require greater focus on the present, and modesty about what healthcare is capable of. It would also be more responsive to screening services as complex socio-technical systems, and the layers of meaning these systems rely on and invoke. By connecting practices to their epistemic, cultural and ontological roots and working across social science, ethics and public health literatures, I hope to have made some contribution to understanding cancer screening services, and so creating space to make them more justified and legitimate.
Footnotes
Declaration of conflicting interests
The author declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.
Funding
The author disclosed receipt of the following financial support for the research, authorship and/or publication of this article: This work was funded by NHMRC grant 1104136.