I am agitating for philosophers of medicine and philosophically-minded clinicians to lead the charge against the term “effective treatment”. This phrase has become ubiquitous in the medical and philosophical literature. But it is a misguided choice which misleads the public and practitioners alike. In general, I want to promote talk about treatment effects not treatment effectiveness. We should be concerned to know and predict effects of treatments—whether beneficial or harmful. Treatment effectiveness is a shoddy stand-in for treatment effects.
There are two primary pernicious effects of the “effective treatment” phrase. First, it reinforces a false binary between effective and ineffective. It encourages us to overlook effect sizes, which I’d argue are the most important and relevant information in making clinical decisions. Saying that a treatment is ‘effective’ conveys relatively little of the information we want—either as consumers or recommenders of the treatment. Testing for effectiveness sidelines the real important issue in medical research: predicting effect sizes. It also misses out a crucial third component: harmful effects. Effective treatments are not just those treatments which have effects! They are treatments which have beneficial effects. But many treatments turn out to be harmful for some patients, and nearly all treatments have both benefits and harms for some or all patients. As such, we should talk about the effects of a treatment.
Second, it flattens out the landscape of treatment effects. Treatments are usually called effective or ineffective across the broadest possible population (or at least, the broadest population about which we have data). But we should be completely aware that the same treatment can have different effects upon patients with different features (and under different conditions). It is a mistake to call a treatment “effective”, or to base a clinical judgment on that label, where effectiveness is assessed in a wide population. Rather, clinicians and patients alike are concerned with predicting the likely treatment effect upon them. We consistently find that a positive net average treatment effect in a broad population is composed of pockets of patients in which the treatment had a considerable greater effect on the outcome measured (high responders), and others in which the treatment had smaller or no effect (low responders and non-responders). The same goes – and indeed is much better appreciated – for side-effects.
The primary projects of clinical science are predictive, not descriptive. The challenge is to work out how to use past experiences of other patients to predict the likely effects of treatments in subsequent cases. Generalising from population-level effectiveness is the least we can do. To do more, we’ll need to work to predict effect sizes for a range of effects, then build from there to predict how those effect sizes predictably vary according to features of the patient, their condition, and their environment. A nuanced approach to prediction seems incompatible with the blunt instrument of the label ‘effective’.