Preventive and Curative Medical Interventions

Abstract

Medical interventions that cure or prevent medical conditions are central to medicine; and thus, understanding them is central to our understanding of medicine. My purpose in this paper is to explore the conceptual foundations of medicine by providing a singular analysis of the concept of a ‘preventive or curative medical intervention’. Borrowing a general account of prevention from Phil Dowe (2000, 2001), I provide an analysis of prevention, cure, risk reduction, and a preventive or curative intervention, before turning to preventive and curative medical interventions. The resulting counterfactual-mechanistic account holds that preventive and curative medical interventions reduce the probability of a medical condition in an actual population compared to their counterfactual omission, commonly by disrupting an etiological or constitutive mechanism for the condition.

1. Introduction

Medical interventions are as central to medicine as equations are to mathematics. Countless resources are consumed in developing and using medical interventions in the quest to cure and prevent disease. Alex Broadbent (2019) even argues that curing and preventing disease is the goal of medicine, and doctors spend much of their time in pursuit of that aim (Fuller 2021). According to Broadbent, to know what medicine is, one needs to be acquainted with health and disease, but also with medicine’s goals. In short, one cannot understand medicine fully without understanding preventive and curative medical intervention.

Philosophers have devoted substantial attention to health, disease, illness, and other medical conditions, but less attention to the interventions that are used to cure or prevent them or otherwise care for the one afflicted¹. Philosophical analysis of medical intervention is valuable in its own right. It can also potentially help with the resolution of nearby problems. Consider boundary problems about the proper scope of medicine and what doctors are permitted to do. These disputes are sometimes made to turn on whether a condition is a disease (Reznek, 1987; Boorse, 2011), and whether medicine has a disease-centered or ‘pathocentric’ internal morality (Boorse, 2016; Hershenov, 2020). However, those advocating for movements like ‘social medicine’ often aren’t promoting different individual-level health outcomes (even if they are promoting a reorientation towards health equity at a population level), but rather they are promoting intervening on different loci in the etiology of disease: on the social determinants of health (Valles, 2018). Whether such interventions are best seen as medical depends on our concept of a medical intervention. Yet even though our concept of a medical intervention may be tied to health or disease, any disagreement about whether interventions on social determinants are ‘medical interventions’ likely doesn’t revolve around what counts as health or disease but rather about what counts as a target of medical intervention.

While prevention has always been among the goals of Western medicine, in the second half of the twentieth century, a new brand of preventive medicine set its sights on the prevention of noncommunicable diseases in individuals by targeting risk factors, including with new medications (Greene, 2008; Aronowitz, 2015). The new preventive medicine reframed prevention around concepts like ‘disease management’ and ‘risk reduction’. This consequential shift raises questions about the relationship between these new concepts and longstanding ideas about prevention and cure; and whether (and how) they can be reconciled. Risk-reducing medications such as antihypertensives, antidiabetic agents, and cholesterol-lowering medications are among the most commonly used drugs today. But questions remain about whether, for instance, treatment of disease should be given priority over prevention of disease, or vice versa (Faust & Menzel, 2012). Preventive medicine is both lauded as well as criticized as a gateway to medicalization and overtreatment (Godlee, 2005; Olivier, 2017). These problems may depend partly on the ways in which prevention and risk reduction are different from treatment and cure and whether these differences are morally or practically significant.

My purpose in this paper is to explore the conceptual foundations of medicine by providing a singular analysis of the concept of a ‘preventive or curative medical intervention’. Although my analysis is partly prescriptive, it is consistent with concepts of prevention, cure, and risk reduction as they are applied in medicine, as well as paradigm examples of preventive and curative medical interventions. Borrowing a general account of prevention from Phil Dowe (2000, 2001), I provide an analysis of prevention, cure, risk reduction, and a preventive or curative intervention (in section 2), before turning to preventive and curative medical interventions (in section 3). My focus is on what makes a medical intervention preventive or curative and how such an intervention works rather than what makes it a medical intervention (versus, say, a public health intervention). The resulting counterfactual-mechanistic account holds that preventive and curative medical interventions reduce the probability of a medical condition in an actual population compared to their counterfactual omission, commonly by disrupting an etiological or constitutive mechanism for the condition.

2. Preventive and Curative Interventions

My aim is to analyze the concept of a preventive or curative medical intervention. I will approach this objective in a stepwise manner. First, in this section, I will analyze preventive and curative interventions. Then in section 3, I will apply my analysis to medical interventions. I will analyze preventive and curative interventions in even smaller steps, beginning with the concept of prevention, before analyzing cure, risk reduction, and, finally, preventive and curative interventions. My analysis of a preventive or curative intervention is partly prescriptive: I seek to suggest an adequate definition of a preventive or curative intervention for medicine. However, some of the adequacy conditions I use are descriptive: this definition should be consistent with concepts of prevention, cure, and risk reduction as they are used in contemporary medicine, as well as with how we commonly come to learn that an intervention is preventive or curative in medicine. Thus, my analysis of prevention, cure, and risk reduction is mainly descriptive, aiming to reconstruct these concepts as they operate in medicine using important examples.

2.1. Prevention

The concept of prevention is a convenient starting point because it is not an idea unique to medicine but also shows up commonly in other sciences and in everyday contexts, as when a government passes a stimulus bill to prevent an economic recession, or a goaltender blocks a shot on net to prevent a goal. Philosophers have accordingly devoted a fair amount of attention to prevention, typically as an offshoot of research on causation.

There is debate about whether prevention is genuine causation given that, unlike in less controversial cases of causation, in prevention the outcome (a recession, a goal scored) does not occur (Dowe, 2001; Beebee, 2004; Hitchcock, 2007; Faust, 2013). The outcome in prevention is a non-occurrence or absence, which poses a problem for accounts of causation in which the relata are described as positive events or outcomes². However, we can (and will) set aside the question of whether prevention is causation and analyze prevention (as well as cure and risk reduction) on its own terms.

Walsh and Sloman’s (2011) empirical research found that individuals (specifically, undergraduate students) are more likely to attribute prevention to an intervention when that intervention is counterfactually related to the outcome, especially when the intervention interrupts a causal mechanism that would otherwise have generated the outcome, consistent with Phil Dowe’s (2000, 2001) theory of prevention³. While this finding is only suggestive, it points towards an intuitive understanding of prevention that (as we will see) is well suited to preventive medical interventions. First, let’s start with a non-medical example. When the goalie blocked the puck (P), they prevented a goal (D) in the sense that a goal did not occur, but had the goalie not blocked the puck then a goal would have occurred. This ‘would have’ describes a counterfactual or contrary-to-fact scenario in that neither the failure of the goalie to block the puck nor the goal actually occurred. Instead, in the actual world the goalie saved the goal by interfering with the path of the puck on its way from the wrist shot (X) to the net.

Dowe captures this intuitive understanding of prevention as follows.

• PREVENTION: P prevented D if P occurred and D did not, and there occurred an X such that

• (P1) there is a causal interaction between P and the process due to X, and

• (P2) if P had not occurred, X would have caused D (2001, p. 221).

Here, P, D and X refer to events or facts. I will draw on Dowe’s account of prevention not because it is the only plausible account of prevention in philosophy but for two reasons. First, by leaving causation unanalyzed in P1 and P2, Dowe intends for his definition to provide a ‘cross-platform solution’, compatible with a range of theories of causation, which allows us to set aside the difficulty of analyzing causation⁴. Second, Dowe’s analysis includes interactions with processes that are responsible for D, which is central to how we understand medical prevention (as we will see in section 3, where D is a medical condition).

Not just any counterfactual comparison P1 versus P2 will do. Imagine that in the actual scenario P1 the goalie blocked a wrist shot that was sailing wide of the net and a goal did not occur. Imagine also a counterfactual scenario in P2 in which instead of blocking this shot the goalie inadvertently redirected the errant shot into the net. PREVENTION would be satisfied and yet it would not be right to say that the goalie’s blocking the shot in P1 prevented a goal. The counterfactual comparison in PREVENTION requires further specification before it can provide an adequate analysis of prevention. I will not complete that task here, but the literature on counterfactuals and causation offers plausible solutions⁵.

Dowe (2000) provides other definitions, including a definition of prevention by omission. Because I am building towards an account of preventive interventions and not preventive omissions, I will not be concerned with prevention by omission. However, I will follow Dowe’s example and state PREVENTION as a sufficient condition rather than a necessary and sufficient condition in recognition of the problem of preventions of preventions (Dowe, 2000). For instance, imagine that an offensive player tried to assist the wrist shot by screening it, blocking the goalie’s line of sight. Imagine also that the offensive player would have succeeded but for a defender on the other team who shoved the offensive player and caused them to fall. We might say that the defender prevented the goal from being scored (without their intervention, the goalie’s vision would have been obscured by the offensive player and the goalie would not have saved the goal). Such a case of ‘triple prevention’ (prevention by prevention of prevention of prevention) exceeds the analysis above. Yet the relevance of this kind of case to medical interventions is dubious. A medical analogy would be a medication reminder (such as an alert on one’s phone) that prevents forgetfulness from preventing one from taking a prophylactic antibiotic that will prevent an infection. However, when we think about ‘preventive medical interventions’, we usually have the medication in mind rather than the medication reminder. Nonetheless, because at this stage I am analyzing prevention generally and not only preventive medical interventions, I will follow Dowe in providing only a sufficient condition for prevention.

Unlike the concept of risk reduction, which I will discuss in 2.3, the concept of medical prevention is not a technical one provided to medicine by epidemiology or any other science. Rather, medicine applies the everyday sense of prevention to medical interventions. For instance, Clarke writes: “The aim of preventive medicine is the absence of disease, either by preventing the occurrence of a disease or by halting a disease and averting resulting complications after its onset…Preventive medicine includes all measures which limit progression of disease at any stage of its course” (1974, p. 65). By interacting with and halting the progression of disease, prevention results in the absence of disease (or its complications). For instance, in patients with severe forms of familial hypercholesterolemia (FH), an inherited disease in which blood cholesterol levels are abnormally high, the high cholesterol results in a heart attack at an early age in the absence of treatment (Defesche et al., 2017). A statin is therefore typically prescribed. Statins interact with atherosclerosis, the process through which high cholesterol causes heart attacks, by lowering blood cholesterol levels so that in at least some patients an early-age heart attack does not occur. Given that a heart attack would have occurred otherwise in at least some of them, according to PREVENTION the statin prevented a heart attack for those patients.

Two problems for any analysis of prevention are ‘over-prevention’ and ‘preventive pre-emption’ (Dowe, 2000), which are analogous to the classic problems of causal overdetermination and causal pre-emption, respectively (Lewis, 1973). In over-prevention and preventive pre-emption, intervention 1, used on its own, would satisfy PREVENTION, but a second intervention is also used that confounds the analysis. In over-prevention, intervention 2 on its own would also satisfy PREVENTION, so adding intervention 1 does not make a difference to D’s occurrence. In preventive pre-emption, intervention 2 will ensure the non-occurrence of D if and only if intervention 1 does not do its job, so that, regardless of what intervention 1 does, D will not occur. It is questionable whether over-prevention or preventive pre-emption are genuine prevention, whether we should say that intervention 1 prevented D. Neither case seems to satisfy PREVENTION because if intervention 1 were (contrary-to-fact) not used but intervention 2 were still used, D would not occur, thus violating P2.

To illustrate, consider an example of over-prevention in which an individual with FH takes a statin as well as ezetimibe, a drug that lowers blood cholesterol by limiting absorption of dietary cholesterol in the small intestine (Michos et al., 2019). Say that there is a threshold for average blood cholesterol above which this individual will have a heart attack before a certain age, but below which they will not have a heart attack before that age. Say that either the statin or ezetimibe taken on its own would lower blood cholesterol in this individual below the threshold, and that taken together they would also lower blood cholesterol below threshold. But in the absence of at least one of these treatments, blood cholesterol would be above threshold and they would have a heart attack. (I will only consider over-prevention here, but we could modify the example so that the statin only lowers blood cholesterol if the ezetimibe does not, in order to confront an example of preventive pre-emption.) Now say that the individual takes both treatments, and as expected their average blood cholesterol is below threshold and they do not have a heart attack. Did the statin prevent a heart attack?

It depends on how the counterfactual is specified. If in the counterfactual scenario the patient took ezetimibe but no statin, then PREVENTION is not satisfied because no heart attack occurs. If instead in the counterfactual scenario the patient took neither ezetimibe nor the statin, then PREVENTION is satisfied because the above-threshold blood cholesterol will cause a heart attack. You might find PREVENTION’s verdict in the second instance (‘the statin prevented a heart attack’) more intuitively reasonable because the statin lowered blood cholesterol enough to bring it from above threshold to below threshold. But you might also find the choice of counterfactual in the first instance more intuitively reasonable because it involves a more minimal alteration to the causal situation compared to ‘removing’ an additional preventive factor, the ezetimibe. How should we resolve this clash of intuitions?

We could lean on a more complete specification of P2 in PREVENTION to solve cases of over-prevention and preventive pre-emption by stipulation. Alternatively, we could index preventive claims to a specific counterfactual comparison. Rather saying that the statin in our example does (or does not) prevent a heart attack, we could say that it prevents a heart attack relative to no statin and no ezetimibe, and that it does not prevent a heart attack relative to ezetimibe alone, which clarifies the intended counterfactual comparison. Or we could let the context implicitly signal the intended comparison. The strategy of explicitly relativizing statements about effectiveness to a specific comparison is often advocated in discussions of counterfactual effect measures in epidemiology (Rothman et al., 2008). While it isn’t often done in medicine when discussing prevention (or cure or risk reduction) in an individual, it often is done when interpreting the results of an epidemiological study of an intervention because such studies involve an actual group comparison that is used (according to the counterfactual approach to epidemiology) to represent a counterfactual group comparison. Applying this practice to individual-level claims could help resolve intuition clashes as well as clarify ambiguous or vague statements about prevention (or cure or risk reduction).

In contrast, Dowe’s solution to over-prevention and preventive pre-emption (which he calls ‘pre-emptive prevention’) is to disjoin P2 above with “there exists a C such that had neither A nor C occurred, x would have caused B or…” (2000, p. 134), where A plays the part of the statin and C plays the part of ezetimibe in my example. With this amendment, PREVENTION delivers the verdict that the statin prevented a heart attack. However, sometimes we may want to say that the statin did not prevent a heart attack (relative to the patient’s current treatment regime) and may even leave this bracketed relativization implicit, indicated by the context. Dowe’s solution overbearingly prohibits us from doing so.

In the preventive medicine literature, a distinction is often made between primary and secondary prevention (and sometimes with tertiary prevention too). Primary prevention is the prevention of the initial occurrence of a disease or medical condition, while secondary prevention is the prevention of the progression of disease or of the development of further manifestations of disease after initial disease has arisen (Clarke, 1974). For instance, a statin may be taken for primary prevention to prevent a heart attack in individuals without cardiovascular disease, or for secondary prevention in individuals with existing cardiovascular disease. Further, in individuals with type 2 diabetes, oral antidiabetic drugs may be used to prevent secondary manifestations or complications of diabetes such as neuropathy or cardiovascular disease, which is one sense in which the expression ‘disease management’ is used in medicine. PREVENTION can be applied to provide a successful counterfactual analysis of each of these examples because the intervention ensures that outcome D – a heart attack or diabetic complication – does not occur (when D would have occurred otherwise).

2.2. Cure

Like prevention, cure is plausibly a causal concept, where to cure means something like ‘to cause to go away’. However, in dissecting cure, I will adopt the same strategy I used in analyzing prevention by analyzing cure on its own terms rather than wondering about whether it is genuine causation. Unlike prevention, cure is a concept less commonly applied beyond the clinic, so my starting point here will be curing disease.

We can start by asking what distinguishes medical cure from natural remission. In natural remission, a disease goes away on its own, while in cure a medical intervention is responsible for a disease going away. Strep throat is a disease that, in people with a normally functioning immune system, generally goes away on its own after a few days. Let’s say that my strep throat is destined to go away by day seven if I don’t take any remedy. Let’s also say that I took a high dose of vitamin C from days three through seven, and my strep throat went away by day seven. Clearly, the vitamin C did not cure my strep throat, my strep throat naturally remitted. Let’s instead say that I took penicillin rather than vitamin C from days three through seven and my strep throat went away by day five. This time, clearly penicillin cured my strep throat. Why is our judgment different in the penicillin case compared to the vitamin C case? To distinguish cure from natural remission, we compare what occurred with the intervention to what would have occurred absent the intervention. With or without vitamin C, my strep throat would have gone away by the same day (day seven) – vitamin C makes no difference. On the other hand, with penicillin my strep throat would have gone away by day five compared to day seven in a counterfactual scenario without penicillin.

Therefore, like prevention, cure is a counterfactual concept. In fact, there are important symmetries between prevention and cure. When a statin (P) prevents a heart attack (D), a heart attack does not occur but would have occurred without the statin. When penicillin (C) cures strep throat (D), strep throat does not persist but would have persisted without the penicillin. Further, in preventing heart attack, a statin interacts with the process due to LDL-cholesterol (X) that causes heart attack (atherosclerosis). Likewise, in curing strep throat, penicillin interacts with the bacterial process of growth and survival due to Streptococcus pyogenes (X) that sustains strep throat. In fact, medical prevention can be seen as extending the paradigmatic medical aim of disease cure to future diseases and the processes responsible for them. Therefore, we can modify our analysis of prevention in providing an analysis of cure⁶.

• CURE: C cured D if C occurred and D ceased occurring, and there occurred an X such that

• (C1) there is a causal interaction between C and either X or the process due to X, and

• (C2) if C had not occurred, X or the process due to X would have continued to sustain D.

CURE differs from PREVENTION in that, in CURE, D is occurring, while in PREVENTION, D has yet to occur. Also, in CURE, X or the process due to X ‘sustains’ D, while in prevention, it causes D. The ‘sustain’ relation in C2 is not unitary. In different diseases, the relevant X or process due to X sustains D in different ways. For instance, in atherosclerosis (D), which is the disease process in which LDL-cholesterol (X) produces a fatty plaque in an artery, the process due to X is identical to D. In an infection (D), a pathogen (X) manifests processes of germ survival and replication and interactions with the immune system that (in some sense) ground D; for example, strep throat is the state of having this infectious process occurring in the throat. In cancer (D), cancer cells (X) undergo processes of uncontrolled survival, replication, invasion, and metastasis that maintain or generate more cancer. Medical cures target these processes: a statin targets LDL-cholesterol, an antibiotic targets processes of pathogen survival and replication, a chemotherapy targets processes of cancer cell survival and replication. While the nature of these metaphysical relationships between the process and the disease is itself an interesting problem, I cannot explore those relationships here. My main point is to illustrate that myriad relationships exist between a disease and the process with which its cure interacts.

A rival analysis of disease cure claims that cure restores the body to a pre-diseased healthful state (Marcum, 2011). One might think that such an analysis follows logically on so-called negative conceptions of health in which health is merely the absence of disease (Boorse 1977): if cure eliminates disease, and health is the absence of disease, does cure not then also restore health? Even setting aside cases in which the individual has multiple concurrent diseases, the answer is ‘no’, because as James Krueger notes (2015), in eliminating disease D a curative intervention could compromise a bodily function or fail to restore a compromised bodily function or substitute for a bodily function that the body performed pre-disease (see also Broadbent (2019)). An example in which a cure compromises a bodily function or fails to restore a compromised bodily function is when curative surgery for cancer removes part of an organ, thus leaving the patient with decreased healthful functioning compared to their health before the cancer. An example in which a cure substitutes for a bodily function is replacement surgery, in which cure replaces a failing organ or heart valve or articular joint with an exogenous one – such interventions do not restore the function of body parts but rather swap those body parts out. Thus, the link between cure of disease and restoration of health is far from conceptually necessary; and is tenuous in actual practice.

Where do medical treatments that require continuous or recurring intervention land? In vitamin deficiencies, treating the deficiency usually requires supplying the deficient vitamin through diet, supplementation, or environmental exposure more than once and often indefinitely. We are happy to say that continuously taking vitamin C or vitamin B-1 cures scurvy or beriberi. However, we do not say that continuously taking exogenous insulin to eliminate hyperglycemia cures type 1 diabetes or that continuously taking an acid-reducing medication cures gastroesophageal reflux disease (GERD), even though the intervention may completely eliminate hyperglycemia or heartburn.

The key to understanding this apparent inconsistency is to probe the nature of the condition D that is (or is not) being cured. Because type 1 diabetes is a chronic condition that persists even when blood sugar levels are normal, it is best understood as a disposition towards hyperglycemia rather than hyperglycemia itself; and likewise, GERD is best understood as a disposition towards acid reflux and heartburn rather than heartburn itself (Fuller, 2018). These dispositions depend on anatomical defects (lack of insulin-producing pancreatic cells, a weak lower esophageal sphincter) that are not eliminated by insulin or acid reducers; thus, those interventions do not cure the disease. In comparison, the fact that we can cure scurvy with vitamin C and beriberi with vitamin B-1 suggests that these diseases consist in their deficiency.

Using my analysis of CURE, we could say that insulin cures hyperglycemia and that acid-reducing medications cure heartburn, but we are more likely to say that insulin ‘controls’ hyperglycemia and that acid-reducing medications ‘relieve’ heartburn. These concepts of control and symptom relief can be given the same analysis as CURE, with the caveat that they are typically applied to manifestations of disease rather than to disease.

2.3. Risk Reduction

In the new preventive medicine, many medical interventions are viewed as ‘reducing the risk’ of some undesirable outcome such as a disease. Interventions like statins that are used for prevention are usually the ones described as risk-reducing, but we can in principle apply the term to curative interventions too. In this subsection, I will analyze risk reduction along the same lines as prevention and cure.

While concepts of medical risk have variable meaning in epidemiology and medicine, one way to understand ‘the risk of’ some medical condition is as expressing a probability (Stegenga & Sprenger, 2017; Fuller & Flores, 2015). When we say that a statin reduces a certain patient’s risk of heart attack over ten years from 0.05 to 0.04, these numbers are best interpreted as the probability that the patient will have a heart attack untreated versus treated. Likewise, when an epidemiologic study finds that a statin reduces the risk of heart attack from 0.05 to 0.04 in a population, these figures can be interpreted as the probability of a heart attack in the population without statin treatment versus with treatment. In fact, an intervention is typically deemed to reduce risk in patients only after it has been found to lower the probability of some outcome in an epidemiologic study like a clinical trial.

‘Risk reduction’ implies a comparison between two levels of risk. An individual with hypertension might have a certain risk of a stroke, which is then ‘reduced’ if their blood pressure goes down. Let’s say that the individual began taking an antihypertensive before the reduction in blood pressure occurred and we want to know whether the antihypertensive reduced their risk (rather than, say, the concomitant change in diet that they implemented). In order to attribute the change in risk to the antihypertensive (rather than the dietary change), we must determine what the individual’s risk would have been in some counterfactual scenario without the antihypertensive, analogous to the problem we faced in attributing cure to penicillin for strep throat. That is why doctors typically infer that an intervention reduces risk in an individual by inference from the results of an epidemiologic study in which population risk was found to be lower in a treated group compared to a comparable untreated group. On the popular counterfactual approach to epidemiological inference in modern epidemiology (Rothman et al., 2008), we can attribute risk reduction to the intervention in the treated group if the control group accurately estimates what the counterfactual risk in the treated group would have been if the treated group were untreated.

When an intervention reduces the probability of some outcome in a population, this finding is consistent with the intervention acting either deterministically or indeterministically in the individuals that comprise that population. For instance, a statin might prevent a heart attack in one out of 100 people, thus accounting for a risk reduction of 0.01 in the population (Chou et al., 2016). If every time any intervention reduced the risk of an outcome in a population it acted deterministically in individuals in this manner, we would not need any additional concept to account for examples of population risk reduction beyond PREVENTION and CURE. However, if we want to allow that some medical interventions might lower the probability of an outcome in an individual from an untreated probability of less than 1.0, PREVENTION and CURE will fail to capture these cases because P2 or C2 will not be satisfied (it is not true that D will definitely occur without intervention). In fact, if a statin lowers an individual’s probability of heart attack from an untreated probability of 0.05⁷, then D probably won’t occur in the absence of intervention.

To allow for the possibility that some risk-reducing interventions lower an individual’s risk (and not just trivially from 1.0 to 0), I provide the following analysis of a risk-reducing intervention:

• RISK REDUCTION: R reduced the risk of D if R occurred and the probability of D was p, and there occurred an X such that

• (R1) there is a causal interaction between R and either X or the process due to X, and

• (R2) if R had not occurred, X or the process due to X would have fixed a probability of D greater than p.

I intend for this analysis to apply to risk reduction at the individual level as well as at the population level. To allow RISK REDUCTION to apply to either a preventive or curative intervention, I have used the vague expression ‘fixed a probability of D’ in R2 to capture the diversity of relationships between X and D covered jointly by PREVENTION and CURE. For instance, the process due to X could fix the probability of D by causing D with a certain probability (as in LDL-cholesterol causing a heart attack D), or the process due to X could fix the probability of D by being identical with D and by persisting with a certain probability (as in the atherogenic process due to LDL-cholesterol being identical with atherosclerosis D). Again, when measuring risk reduction in a population, epidemiologists often index risk reduction claims to a specific comparison: a diuretic antihypertensive might reduce the risk of stroke compared to no antihypertensive, yet it might not reduce the risk of stroke compared to a calcium channel blocker.

2.4. Preventive and Curative Interventions

Having established the meaning of prevention, cure, and risk reduction, it might seem that we can define a preventive or curative intervention as one that satisfies PREVENTION, CURE or RISK REDUCTION. While this would be a reasonable solution, it would not tell us for whom this disjunction must be satisfied, and it is less elegant than the more parsimonious solution I will offer.

To allow for the possibility of indeterministic prevention, which is not permitted by PREVENTION, we can define a preventive intervention as one that prevents D or that reduces the risk of future D. Similarly, to provide for indeterministic cure, which is unrecognized by CURE, we could define a curative intervention as one that cures D or that reduces the risk of D’s persistence. However, a preventive or curative intervention is typically identified after it has been found to reduce the risk of D in a population comparison. As I already alluded in the previous subsection, such a conclusion implies either that some individuals in the population satisfied one of PREVENTION or CURE, or that some individuals satisfied RISK REDUCTION, or that some individuals satisfied RISK REDUCTION while other individuals satisfied one of PREVENTION or CURE. In other words, RISK REDUCTION, applied to a population, captures the myriad ways an intervention might be preventive or curative on a sufficiently broad understanding of ‘preventive’ and ‘curative’⁸. Thus, I will define a preventive or curative intervention as follows.

• INTERVENTION: I_X is a preventive or curative intervention if and only if I_X satisfies

• RISK REDUCTION for some actual population.

Given the analytic similarities between prevention and cure, preventive and curative interventions can be analyzed together. The main difference is that a preventive intervention reduces the risk of future D in some actual population, while a curative intervention reduces the risk of D’s persistence in some actual population. Note that the same intervention might not satisfy INTERVENTION for a different population. According to INTERVENTION, I_X is a preventive or curative intervention as long as there is some population in which that intervention is preventive or curative, however small or obscure that population might be. So long as there is at least one individual for whom I_X satisfies PREVENTION, CURE, or RISK REDUCTION, I_X is a preventive or curative intervention because there will be some populations containing that individual (including a one-person population containing only that individual) for which I_X satisfies INTERVENTION. INTERVENTION, open-minded as it is, even allows that I_X can reduce the risk of D in some population while also raising the risk in the vast majority of populations.

One might therefore worry that INTERVENTION is too weak, too easy to satisfy. Why not instead require that I_X satisfy INTERVENTION for all populations, or at least for most populations, before we call I_X a preventive/curative intervention? There are several drawbacks to this suggestion. First, this proposed convention is not descriptively accurate. Even if we add a rider to the suggestion stating that INTERVENTION must be satisfied for all or most populations for which the intervention is indicated, our revised definition would not be faithful to linguistic practices in medicine. Clinicians consider a drug or procedure to be a medical intervention even if they think it works in few individuals. In the very least, clinicians are happy to consider a drug or procedure to be a medical intervention when they do not know whether it works in most or all populations because it has not been studied in most or all populations (which most contemporary medical interventions have not). In primary prevention, a statin will only reduce the risk of fatal or nonfatal heart attack or stroke by less than 0.01 in the population consisting of all individuals who are generally screened for treatment; namely, adults 40 years of age or older (Chou et al., 2016). That is compatible with the statin preventing a heart attack in very few individuals, in which case it will not reduce the risk of heart attack in most populations because most populations will not contain those individuals. In short, clinicians consider something to be a medical intervention if it satisfies a definition like INTERVENTION rather than a definition requiring it to reduce risk in most or all populations.

Second, medicine ought not adopt this proposed linguistic convention, for two reasons. The first reason is that while it seems clear that a drug or procedure should at minimum reduce risk in some population before we call it a preventive or curative intervention, it is arbitrary to specify any given proportion of populations for which it should reduce risk, whether that proportion is greater than 50% (‘most populations’) or 100% (‘all populations’). A great many of the possible populations we could define will be hackneyed (for instance, one population might consist of you, me and Julius Caesar), so we should attach no special significance to a drug that lowers risk in most or all of them. The second reason why medicine ought not accept this proposed amendment is that clinicians should separate the question of whether a drug or procedure is a medical intervention from the question of how effective it is. Yet, requiring that the drug or procedure work in most or all populations partly collapses this distinction because the number of populations in which a drug/device works is relevant to how effective that drug/device is. The effectiveness of a medical intervention is better expressed by quantifying the effect size in specific populations rather than by partially baking a quantification of effectiveness into the concept of an intervention.

Though I won’t provide a conceptual analysis of ‘effectiveness’ with respect to preventive and curative medical interventions⁹, a successful analysis should be consistent with my account of what it means for something to be a preventive or curative intervention. For starters, an effective preventive or curative medical intervention must satisfy INTERVENTION. The effectiveness of a medical intervention is commonly quantified according to the amount of risk reduction it achieves in a population. Based on my analysis in this section, this population-level effectiveness depends on the amount of cure, prevention, and risk reduction the intervention achieves for the individuals in that population. But just as RISK REDUCTION is agnostic as to whether population-level risk reduction is due to individual-level cure/prevention or rather due to individual-level risk reduction (or a combination of both), population-level quantifications of effectiveness must remain agnostic in the absence of further evidence to decide among these possibilities.

INTERVENTION only requires that I_X has some efficacy and makes no assumptions about how efficacious I_X is. However, INTERVENTION does require that I_X reduced risk in an actual population rather than a hypothetical population. The populations in whom it reduced risk might even be purely historical, given that RISK REDUCTION is characterized in the past tense. So, the smallpox vaccine will still be considered a preventive intervention 100 years from now even though there will (presumably) be no population living for whom it reduced the risk of smallpox. It seems reasonable, however, to require that there is some actual population in which a medical intervention reduced the risk of D rather than allow that it might only reduce the risk of D in some possible world because the sense of preventive or curative intervention I am after is the sense in which an intervention is actually (rather than merely hypothetically) preventive or curative – it actually reduces risk.

Finally, the conceptual similarity between prevention and cure and the common understanding of a preventive or curative medical intervention provided in this section makes it less plausible that we should attach moral significance to whether an intervention is preventive versus curative. What may matter more in priority-setting in healthcare is the amount of risk and risk reduction expected in a population, the size of the population (Rose, 1992), and how the population’s risk is distributed among its individuals (Daniels, 2012; John, 2014).

In summary, a preventive or curative intervention reduces the risk of D in some population by reducing the risk in some individuals and/or by preventing or curing D in some individuals. All these concepts of prevention, cure, risk reduction, and an intervention are best given a similar counterfactual analysis, which I provided in this section.

3. Preventive and Curative Medical Interventions

We have arrived at a definition of a preventive or curative intervention but have not yet determined what makes for a preventive or curative medical intervention. In this section, I will apply the concepts from the previous section to medical conditions and their mechanisms. My main aim is to elucidate through paradigm examples how preventive and curative medical interventions typically work.

The concept of prevention can be used outside of medicine to describe the prevention of economic recessions or hockey goals. Likewise, the concepts of cure, risk reduction, and a preventive or curative intervention as analyzed in the previous section could in principle be applied beyond medicine. So, what makes a particular intervention a medical intervention as opposed to an economic intervention? An obvious answer is that a preventive or curative medical intervention prevents, cures, or reduces the risk of a ‘disease’ or ‘medical condition’. Jacob Stegenga (2018) argues that a curative medical intervention targets the constitutive causal basis of a disease (while an effective medical intervention more generally targets either the constitutive causal basis or the harms of a disease). Most of Stegenga’s analysis probes the concept of disease as well as how its conceptual elements (for Stegenga, ‘constitutive causal basis’ plus ‘harm’) map onto the goals of medical treatment (‘cure’ and ‘care’, respectively). In contrast, in the previous section I concentrated on the counterfactual notion of what it means to be a curative or preventive medical intervention; and in this section, I’ll focus on the mechanics of medical prevention and cure.

What distinguishes economic prevention from medical prevention is the target of prevention: in economics the target is the economy, while in medicine, the target is a medical condition. All we must do is add the requirement that D is a medical condition to PREVENTION, CURE, and RISK REDUCTION and we have a proper analysis of medical prevention, medical cure, medical risk reduction, and a preventive or curative medical intervention. A statin and an oral antidiabetic drug are preventive medical interventions because they prevent or reduce the risk of medical conditions like heart attacks; and penicillin and surgery are curative medical interventions because they cure or reduce the risk of the persistence of infection and cancer, respectively (they are also preventive medical interventions because they can be used prophylactically to prevent infection or cancer). While I think this answer is right, it is not without difficulties, which I will sort out in the remainder of this section.

The first difficulty is that it is not obvious what a medical condition is. We have analyzed the murky concept of a preventive or curative medical intervention using the equally murky concept of a medical condition. Analyzing a concept like ‘disease’ or ‘medical condition’ is a formidable (and frequently faced) problem of its own that I will not attempt here. Rather, I will examine medical interventions through representative examples. In prevention, I will take two of the most commonly used classes of preventive medications: statins and antidiabetic agents. Meanwhile, the two paradigmatic examples of medical cures are cures for infections and cures for cancer, so those interventions will serve as my curative examples. I will leave it to the ‘concepts of health and disease’ literature in philosophy of medicine to provide a conceptual analysis of a medical condition. Note, however, that I mean for the concept of a medical condition to be wider than the concept of a disease, and to include not only pathological conditions but also non-pathological conditions like pregnancy (for which contraceptives are plausibly ‘preventive medical interventions’). Note also that the requirement that D is a medical condition is a requirement on preventive and curative medical interventions specifically and might not extend to all medical interventions. Whether bona fide medical interventions like acetaminophen relieve a medical condition depends on whether (medically important) symptoms like pain are ‘medical conditions’.

Even after clarifying that D must be a medical condition, the resulting analysis of a preventive or curative medical intervention may be incomplete. It may provide only a necessary condition for a preventive or curative medical intervention (the intervention must reduce the risk of a medical condition in some actual population). This criterion might not be sufficient. Consider bike helmets and economic relief programs. Bike helmets reduce the risk of traumatic brain injuries, while economic relief programs reduce the risk of various medical conditions when those conditions result from lack of economic resources. Yet we might not consider bike helmets or economic relief programs to be medical interventions. Perhaps they are instead ‘public health interventions’ and ‘economic interventions’. I will not weigh in on whether we do or should consider examples like these to be medical interventions, and thus my analysis of a preventive or curative medical intervention may remain incomplete. My main concern in this paper is understanding why and how preventive and curative medical interventions are preventive and curative rather than mapping their boundaries on the ‘medical’ axis.

Clarifying that a particular preventive or curative medical intervention is one that is counterfactually related to some medical condition in the way I explained in the previous section provides only minimal understanding of how any such intervention works; all it says is that a preventive or curative intervention reduces the risk of the medical condition in some way. However, my analysis of prevention, cure, and risk reduction suggests an avenue for further clarification. Each of these concepts holds that the intervention causally interacts with X or the process due to X, which would otherwise cause or sustain D. As I will now illustrate through examples, when D is a medical condition, the relevant process is a mechanism in the ‘new mechanist’s’ sense, consisting in entities and activities organized such that they are responsible for the phenomenon D (Illari & Williamson, 2012)¹⁰. Mechanisms are processes in the colloquial sense of ‘process’ as an ordered sequence of occurrences (Dammann, 2020). Therefore, we can profitably understand preventative and curative medical interventions by examining the ways that they interact with mechanisms.

First, we should distinguish two distinct relationships or sets of relationships between the mechanism and the medical condition D. In PREVENTION, the mechanism causes the medical condition, while in CURE the mechanism (or some factor X) sustains the medical condition. To distinguish these two different relations, I will borrow a distinction from Carl Craver (2007) and Wesley Salmon (1984) between etiological and constitutive explanations. In etiological explanations, a phenomenon is explained using the mechanism that causes it. I will refer to the mechanism that causes a phenomenon as the phenomenon’s ‘etiological mechanism’. In constitutive explanations, a phenomenon is explained using the mechanism ‘underlying the phenomenon’ (more on this expression below). I will refer to the mechanism underlying a phenomenon as the phenomenon’s ‘constitutive mechanism’. In brief, in PREVENTION, a preventive intervention causally interacts with a medical condition’s etiological mechanism, while in CURE, a curative intervention causally interacts with a medical condition’s constitutive mechanism (or with some factor X)¹¹.

Typically, for a preventive intervention to prevent or reduce the risk of a medical condition there must be some etiological mechanism with which the intervention causally interacts because rarely do preventive interventions interact directly with a medical condition – ‘directly’ meaning (roughly) without any causes mediating between the intervention and the medical condition. Rather, the intervention will more directly interact with some other cause, an etiological factor – and that etiological factor will typically not be a direct cause of the medical condition either. Therefore, if the intervention is to have any influence on the medical condition, there must be some causal mechanism – an etiological mechanism – connecting the etiological factor to the medical condition because if there are no causal intermediates linking the etiological factor to the medical condition and the etiological factor is not a direct cause of the medical condition, then there is no causal connection at all between the etiological factor and the condition; and, intuitively, if there is no causal connection between an etiological factor and the condition, then interventions on that etiological factor cannot have any preventive influence on the condition¹².

In what way must a preventive intervention ‘causally interact’ with an etiological mechanism in order to prevent or reduce the risk of the medical condition in an individual? Roughly speaking, the answer is that the intervention must disrupt the mechanism. What it means to ‘disrupt a mechanism’ depends on whether the intervention is deterministic or indeterministic – whether we are dealing with PREVENTION or RISK REDUCTION. I will start with deterministic prevention. Daniel Steel offers the useful idea of a ‘disrupting factor’, which “can be thought of as a switch that, when set to a particular position, breaks the mechanism connecting the cause and the effect” (Steel, 2008, p. 60). Assume that all causes in the mechanism can be described dichotomously as either the occurrence or non-occurrence of that cause. Consider a simple abstract mechanism: (the occurrence of) X causes (the occurrence of) Y, and (the occurrence of) Y causes (the occurrence of) Z. Moreover: without X, Y does not occur; and without Y, Z does not occur. In prevention, Z is equivalent to the medical condition D (in cure, X or X’s mechanism sustains D). Preventive intervention P can prevent Z by stopping X from causing Y, thus disrupting the mechanism at Y.

To take a real example, a statin can prevent a heart attack by disrupting atherosclerosis, the etiological mechanism for heart attack. Statins lower blood LDL-cholesterol (among other possible effects (Wolfrum et al., 2003)). If we describe the level of average blood LDL-cholesterol dichotomously as being either ‘above threshold’ or ‘not above threshold’ for significant LDL-cholesterol deposits in a coronary artery, then the (greatly simplified) etiological mechanism is as follows: biological and behavioral factors cause above-threshold LDL-cholesterol, which causes pathological LDL-cholesterol deposits in a coronary artery, which causes an inflammatory cascade, which produces an unstable arterial plaque, which causes plaque rupture or erosion, which causes blood clot formation, which causes arterial occlusion, which causes myocardial ischemia, which causes myocardial infarction (Boren et al., 2020). A statin disrupts atherosclerosis by disrupting the connection between biological and behavioral factors (such as diet and genetics) and above-threshold LDL cholesterol, artificially bringing blood LDL-cholesterol below the threshold for significant LDL-cholesterol deposits. (It does so more specifically by inhibiting an enzyme, HMG-CoA reductase, that is required for LDL-cholesterol biosynthesis in the liver so that the liver meets its cholesterol requirements by upregulating LDL receptor expression and snatching up LDL-cholesterol from the blood (Michos et al., 2019).) Without above-threshold LDL-cholesterol, the atherosclerosis chain is broken, and a heart attack does not occur.

In contrast, indeterministic prevention occurs when a preventive intervention reduces the risk of a medical condition in an individual. In our abstract X-Y-Z mechanism, P might lower the probability of Z by lowering the probability of Y. If the probability of Z in the absence of P is less than 1.0, then P2 of PREVENTION is not satisfied, and while P reduces the risk of Z it cannot prevent Z (it is better represented by the R in RISK REDUCTION). In our atherosclerosis mechanism, a statin may reduce blood LDL-cholesterol below threshold, but if the probability that LDL-cholesterol would be above threshold (and/or the probability of downstream stages in the mechanism) is less than 1.0 for an individual in the absence of intervention, then the statin satisfies RISK REDUCTION rather than PREVENTION.

In other examples, P might cause a mechanism stage to occur or might raise its probability. Imagine that now X inhibits Y, which in turn inhibits Z. Normally, X occurs, which causes the non-occurrence of Y, which results in the occurrence of Z. P might this time cause Y to occur, which in turn causes the non-occurrence of Z. In so doing, P disrupts the normal mechanism described above and prevents Z. Based on all these examples, to disrupt a mechanism of disease is to cause or inhibit (or raise or lower the probability) of any of the mechanism’s stages, such that the mechanism no longer causes or sustains the disease and the disease is prevented or cured (or its probability is reduced). A preventive intervention disrupts an etiological mechanism in this way.

This account of mechanism disruption applies equally to disease management and secondary prevention. Oral antidiabetic agents manage diabetes by disrupting etiological mechanisms generating diabetic complications like neuropathy and cardiovascular disease. These complications are ultimately the result of hyperglycemia. Although different antidiabetic agents act in different ways, they all promote a lowering of blood sugar. For example, an SGLT2 inhibitor inhibits the SGLT2 glucose transporter in the proximal tubules of the kidneys. Normally, SGLT2 contributes to hyperglycemia in diabetes by reabsorbing glucose from the lumen of the proximal tubule into the bloodstream. By inhibiting SGLT2, the drug disrupts the connection between SGLT2 expression and SGLT2-mediated glucose reabsorption, with a net lowering effect on blood sugar (Hattersley & Thorens, 2015).

The statin and SGLT2 inhibitor examples illustrate that the medical intervention need not disrupt the mechanism responsible for disease at a malfunctioning (meaning abnormally functioning) pathway and restore normal functioning; the intervention need only disrupt the overall pathological mechanism at some node and promote the absence of the medical condition¹³. This might involve intervening on a pathway within the overall mechanism that is functioning normally, or as it would in the absence of disease. Statin therapy works by disrupting the normally functioning cholesterol biosynthesis pathway. However, through its causal connection to blood LDL-cholesterol, this pathway nonetheless is part of the overall etiological mechanism for heart attack, as indicated by the fact that disrupting cholesterol biosynthesis decreases blood LDL-cholesterol and thus the risk of heart attack. Similarly, in type 2 diabetes, resistance to the effects of insulin in organs like liver and muscle results from a malfunction in insulin signaling. However, rather than the insulin resistance pathway, an SGLT2 inhibitor disrupts the normally functioning pathway of glucose reabsorption in the kidneys. Nonetheless, this renal mechanism is part of the etiological mechanism involving hyperglycemia because it helps maintain high blood sugar, as indicated by the fact that disrupting the reabsorption pathway reduces hyperglycemia.

Thus, a medical intervention does not always disrupt a malfunctioning pathway; it may instead disrupt a normally functioning pathway. However, so long as the normally functioning pathway is part of the etiologic or constitutive mechanism of disease (or medical condition), disrupting it has the potential to mitigate the disease. A pathway is part of the mechanism of disease when it is partly responsible for the disease phenomenon, helping to cause or sustain it. Here, I am following a standard convention in the mechanisms literature in philosophy of science that considers a mechanism to be ‘functionally individuated’ (Glennan, 1996). In other words, we identify a mechanism’s parts and distinguish them from the mechanism’s ‘surroundings’ according to whether they contribute to the demarcating phenomenon – in this case, a particular medical condition. By providing a check on the number of LDL receptors in the liver available to mop up excess blood LDL-cholesterol, cholesterol biosynthesis helps maintain high LDL-cholesterol and atherosclerosis. By returning glucose filtered into the renal tubule back into the bloodstream, glucose reabsorption in the kidney helps maintain hyperglycemia and the mechanisms through which hyperglycemia manifests in diabetic complications.

Turning now to medical cure, Stegenga argues that a medical cure targets the “constitutive causal basis of a disease”, by which he means “states that are constitutive of disease; a disease’s basis; the pathophysiological causes of patient-level symptoms”, as opposed to the “causal etiology of a disease” (2018, p. 26)¹⁴. In CURE, the constitutive basis of the disease was labeled ‘X or the process due to X’, where X is a factor that sustains D or that generates a process that sustains D. Using our terminology in this section, a process that sustains D is a constitutive mechanism of disease (or medical condition). I noted that for different conditions, X or the constitutive mechanism ‘sustains’ the medical condition in different ways. Again, I will not examine the different metaphysical relations captured by this general term. Thus, I will not provide a general metaphysical theory of what it means for a constitutive mechanism to be ‘constitutive of’ or to ‘underly’ the phenomenon (but see: Craver, 2007; Kaiser & Krickel, 2017; Kastner & Andersen, 2018).

Similar to preventive interventions, one way in which a curative intervention can succeed is by disrupting a mechanism; in this case, a constitutive mechanism that sustains or underlies a medical condition. With preventive interventions, disruption worked by severing a link in a causal chain that would have otherwise caused the medical condition. But the relationship between a constitutive mechanism and a medical condition is generally noncausal (with the possible exception of cancer, in which by proliferating and spreading, cancer cells generate more cancer). Nonetheless, the ‘sustain’ relationship is intimate enough that disrupting the constitutive mechanism cures or reduces the risk of the medical condition. For example, because atherosclerosis is identical to its constitutive mechanism, disrupting the mechanism will cure or reduce the risk of atherosclerosis (if only temporarily).

In infections, disrupting constitutive mechanisms that ensure pathogen survival and/or replication often cures the disease. For example, penicillin inhibits the enzyme transpeptidase, which is found in Gram positive bacteria (Wise & Park, 1965; Tipper & Strominger, 1965). Transpeptidase catalyzes the crosslinking of molecules of the glycoprotein peptidoglycan, which strengthens the bacterial call wall, allowing the cell wall to provide the bacteria with the structural rigidity needed for survival. By inhibiting transpeptidase, penicillin weakens the cell wall, leading to the death of Gram-positive bacteria like Streptococcus pyogenes, which is responsible for strep throat. By disrupting this constitutive mechanism for strep throat, penicillin cures the strep.

In cancer, mechanisms of abnormal cancer cell survival, replication, invasion, and metastasis maintain existing cancer cells and generate more cancer cells at the primary site as well as distant sites. Cancer chemotherapies often work by disrupting constitutive mechanisms of cancer cell survival and replication. For example, vincristine is a chemotherapy in use since the 1960s and is still a standard treatment for childhood acute lymphoblastic leukemia (ALL). It binds to tubulin (a protein comprising microtubules in the cell), preventing microtubule formation (Himes et al., 1976). In dividing cells, this prevents the formation of the mitotic spindle, a structure necessary for mitosis. Thus, vincristine disrupts the mitotic mechanism required for cell proliferation. Combined therapy with multiple chemotherapies, including vincristine (as well as radiation and targeted therapy), cures childhood ALL in many cases (PDQ, 2021).

However, many systemic therapies for other cancers are used with the goals of improving the success of curative surgery, or – bespeaking the dominance of the new preventive medicine – preventing progression of, recurrence of, or death from cancer. For instance, according to a recent review, “For nonmetastatic breast cancer, the main goals of therapy are eradicating tumor from the breast and regional lymph nodes and preventing metastatic recurrence” (Waks & Winer, 2019, p. 290). Many initial cancer interventions are not described as achieving cure but rather remission from cancer (freedom from cancer signs and symptoms), while other times interventions explicitly aim at preventing recurrence. Whether or not remission constitutes cure according to our definition depends on how cancer is conceived; in particular, whether the persistence of undetectable cancer cells in a patient in remission constitutes having cancer. If it does, then cancer therapies that promote prolonged remission are better understood as preventing recurrence (as the return of clinically apparent cancer) rather than achieving cure.

While many medical cures work by disrupting a constitutive mechanism, they do not all work this way. For example, often surgeries are too blunt to reasonably be described as disrupting a mechanism at one or even several stages; ironically, surgeries are typically not ‘surgical’ or targeted in the way required of a disrupting factor. Surgeries instead remove the diseased body part, including the mechanism or factor that sustains the medical condition, thereby curing the medical condition. Rather than disrupting mechanisms of cancer cell survival and replication by intervening on some node in a molecular pathway, a surgical oncologist will remove the cancer cells from the body. The same logic applies to surgeries that prevent cancer by removing breasts, colon, or other body parts in people at high risk. Similarly, in replacing a body part such as a failing heart in severe heart failure, a surgeon cures the medical condition without disrupting a mechanism at a precise node. These exceptions to the paradigm of mechanism disruption are why PREVENTION, CURE, and RISK REDUCTION require that the intervention causally interacts with X or X’s mechanism but they do not require that the intervention is a mechanism disruptor.

My analysis of mechanism disruption could explain why the scope of preventive medicine is potentially so wide while the scope of curative medicine may be less so¹⁵. An etiological mechanism could stretch outward far and wide from the locus of disease to encompass its environmental and social determinants. Without further restriction on the concept of a preventive medical intervention¹⁶, all these nodes potentially fall under the purview of preventive medicine. In comparison, because a disease is a bodily state (or process or object), its constitutive mechanism is also plausibly wholly within the body or at most includes only those non-bodily causes that overlap temporally with the occurrence of the disease, thus ruling out temporally and causally distant factors as relevant for curative medical intervention¹⁷.

In summary, preventive and curative medical interventions prevent, cure, or reduce the risk of a medical condition. Preventive interventions often do so by disrupting an etiological mechanism that causes the condition, while curative interventions often do so by disrupting a constitutive mechanism that sustains the condition.

4. Conclusion

I proposed a counterfactual-mechanistic account in which a preventive or curative medical intervention reduces the risk of a medical condition in some actual population, commonly by disrupting a mechanism for that condition. By ‘reduce the risk of the medical condition in the population’, I mean that the probability of the medical condition is lower compared to a counterfactual scenario in which the intervention is omitted. It can achieve this probability-lowering by preventing, curing, and/or reducing the risk of the condition for individuals in that population, where these concepts of prevention, cure, and risk reduction are given an analogous counterfactual interpretation. Preventive medical interventions achieve this feat by causally interacting with an etiological mechanism, while curative medical interventions do so by causally interacting with a constitutive mechanism.