Defining disease in the context of overdiagnosis

How to define overdiagnosis is under dispute. Welch et al. (2011, xiv) define it as occurring “when individuals are diagnosed with conditions that will never cause symptoms or death”. Carter et al define it by saying “An (asymptomatic) person is diagnosed with a condition; that diagnosis does not produce a net benefit for that person” (2015). The reference to net benefit in some definitions may be problematic since ‘diagnoses that do not result in net benefit’ would capture a broader range of cases where there are unintended iatrogenic harms (see Rogers and Mintzker 2016). Some also allow that overdiagnosis can happen to patients with symptoms. We clarify this rough statement in the discussion in the “Overdiagnosis” section.

The term has sometimes been used in a broad sense, to include a set of related problems such as medicalisation, and overtreatment, as well as in the narrower sense we focus on here. See Carter et al (2015); Moynihan, Doust, and Henry (2012).

We limit our discussion to physical disease, as psychiatric disease raises further complexities that are outside our current scope.

For example, while there is some evidence indicating that mammography screening programs can reduce deaths from breast cancer when targeted at women aged 50–69 (Barratt 2015), this is not the case for breast cancer screening that involves women younger that 50 (Autier 2015) or older than 70 (Glas et al 2014).

These are not mutually exclusive types as there are cases where both misclassification and maldetection occur, such as pulmonary embolism.

This is perhaps the difference between overdiagnosis and other cases where a diagnosis causes overall net harm to a patient (see note 1): in overdiagnosis, the condition diagnosed seems to be the same as cases where a diagnosis will benefit the patient. This could indicate that it is the epistemic problems that are definitive of overdiagnosis, though we do not argue this here.

Here we are using ‘metaphysical indeterminacy’ to refer to open causal chains from the currently identified abnormal state, and its future progression or non-progression (we ignore the possibility of metaphysical determinism for the purposes of this paper). We note that this differs from the way that Hofmann (2016) uses ‘indeterminacy’ to refer to the social question of whether or not a condition comes to be considered a disease through the process of medicalisation (i.e. whether some conditions are diseases is resolved by social and political processes rather than by appeal to underlying biological processes). Thus we are not here disputing Hofmann’s analysis of the differences between medicalisation and overdiagnosis.

We thank an anonymous reviewer for reminding us of this point.

We use ‘dysfunction’ in Boorse’s naturalistic sense here and throughout the paper.

It also challenges the link to the imperative to intervene that some have emphasised (e.g., Engelhardt 1975). We leave this aside in this section, as we consider it linked to the features of harm and increased risk of harm.

We will ultimately depart from these theories, in arguing below that ‘disease’ is a vague concept that is not classically structured.

This feature of the BST is demonstrated in the ‘one dead cell’ objection: that the BST implies disease is present where there is just one dead cell, since that cell is failing to make its species-typical contribution to the biological goals of the organism, even if there is no noticeable change in overall functioning (Nordenfelt 1987, 28, cited in Wakefield 2014, 656). Boorse accepts this implication for tissues that do not usually involve regular cell death and regeneration (1987, 365, in Wakefield 2014, 656). For tissues that do involve regular cell death and regeneration, he indicates that one dead cell would not be pathological since it is not statistically atypical (1997, 50–51). Correlatively, one might say that if it turns out to be normal to have microscopic thyroid cancers, or other lesions, they are not disease on the BST. However, Boorse also accepts that there are some ‘universal’ diseases, such as tooth decay, and his ways of dealing with those would also apply to universal maldetected cancers. We leave further discussion of this aside since our focus is assessing the relevance of dysfunction to disease, not statistical frequency.

This is not a satisfying answer since on his theory what counts as a functional deficit is supposed to respond only to the standard of statistical frequency (for more detail on this problem, see Schwartz 2007a). In any case, there does not seem any obvious cut-off point at which the dysfunction involved should be thought unusual enough to indicate the presence of disease.

‘Harm’ may of course refer to a range of things, including pain and suffering, incapacity, or presence of symptoms. Though we recognise that there are various ways these things might come apart, we will discuss incapacity, pain and suffering simply as ‘harms’ in what follows for reasons of space.

Wakefield uses ‘disorder’ rather than ‘disease’, but states that the term fulfils a similar role in his account to ‘disease’ in other parts of the debate. We thus alter his terminology here for consistency.

One might also think that the presence of such dysfunction itself constitutes a harm. We leave this aside since it could lead to an elision of the two criteria. Thus we are considering harm as an evaluative and not biological notion.

See Schwartz (2007b, 50–56), for a detailed demonstration of how this assumption has influenced the kinds of definitions that have been offered, and the way they are developed, in the philosophical debate on disease definition.

A more detailed discussion of these matters is available in Walker and Rogers (forthcoming b).

One response to this is to consider all of these cases to be disease, and make distinctions about the severity of that disease. Another is to claim that disease itself could be a matter of degree. We do not seek to resolve here which is the preferable response. For a detailed discussion see Rogers and Walker (forthcoming).

Though we focus on physical disease, our précising definition may be of value in conceptualising psychiatric overdiagnosis.

These parameters could, for instance include specified kinds or amounts of dysfunction, as well as certain minimum levels of risk. How and by whom these parameters would be set is a further issue that we will not engage with here.