Abstract
I first heard ‘diagnosis as an alibi for treatment’ at the third Preventing Overdiagnosis conference, held at the US National Institutes of Health in Bethesda, Maryland in September 2015. The concept comes from the writing of Marshall Marinker in 1994. When prescribing an antibiotic for a child with a simple cough the GP may well write ‘bronchitis’ in the notes. An antibiotic in this case is likely to be an instance of overtreatment: it is unlikely to benefit the child, it may cause harm and add to the scourge of antibiotic resistance. However, the doctor may want to treat with an antibiotic for any of a number of reasons including clinical judgement, extra caution due to the child's medical history, or the insistence of the mother. Marinker explained that writing bronchitis in the notes ‘is not so much the basis for the choice of drug but rather the alibi for it’. As a result, the clinical note is no longer an even-handed statement of symptoms and signs written prior to reaching a decision about treatment but is written in the knowledge of the treatment decided upon and hence – ‘diagnosis is an alibi for treatment’. It might be harsh and unfair to regard this as deliberate falsification; hence, the word ‘alibi’ is a neat fit. Nevertheless, it insidiously corrupts the clinical record.
If there are systematic biases in entries in registries and databases, subsequent analysis of ‘big data’ might result in big mistakes. If we consider primary lung cancer, there are three competing approaches to treatment: (1) surgical resection through a thoracotomy; (2) the same resection but by less invasive videothoracoscopy; or (3) by stereotactic radiotherapy. Large databases have been used to compare outcomes between them. The patients in those treatment groups are of course not alike. Various features of the individual patient and their cancer are considered in the multidisciplinary team meeting, and the enthusiasm of the individual surgeons or clinical oncologists is taken into account and a decision is made. Clearly, that is not a random treatment assignment. Any subsequent comparison is likely to start with multivariable analysis and then a matching process to create groups with similar features for a fair comparison of the outcome. So far so good, but now let us consider this from the perspective of ‘diagnosis as an alibi for treatment’. The decision as to which treatment is to be selected is likely to have been made and treatment maybe completed before the patients’ details are entered into the database. Some data entries are fixed and verifiable such as age and sex. Others should be reasonably objective, but many are amenable to interpretation of emphasis and insidiously corrupt the scientific record. This was shown to add up to a large effect at the time of introduction of risk adjusted reporting of cardiac surgery in the 1990s. For example, the proportion of patients with respiratory dysfunction entered into the database shot up. It allowed doctors to shift a proportion of the observed death rate from the surgeons’ competence to the patients’ comorbidity. This was called ‘gaming’.
To return to lung cancer, if open surgery is seen as the gold standard, there may be feeling that the decision to use radiotherapy or keyhole surgery should be justified for the record. As the data are being entered, the chosen treatment is known, and the entries may be subliminally gamed to lend support to a decision to deviate from the gold standard. Statistical adjustment and propensity matching rely on what was recorded. If that was already subtly different depending on the treatment, these data entries may be a form of ‘alibi for treatment’. The more marginal the difference between treatments and the more variables influence the outcome, the harder it is to distinguish the treatment effect from the noise. That is why at least some surgeons and interventional oncologists see the need for randomised controlled trials to balance not only the obvious and the known variables but unsuspected confounding variables.
The trouble with latching on to an idea such as ‘diagnosis as an alibi for treatment’ is that you might see it wherever you look. So what about the ‘oligometastatic state’? For a patient with blood-borne metastases, the first choice of treatment is generally chemotherapy: the blood will take the chemotherapy to wherever the metastases are. But if there are just a few, or that is to say only a few visible, they may be amenable to surgery or ablation. And so ‘oligometastasis’ was coined by oncologists as a diagnosis to describe a situation in which there are ‘few enough to zap’. That is how the state is defined – by their fewness. There is no biological basis yet shown. The oligometastatic state is variously but arbitrarily defined as ≤5 or ≤3 metastasis. So oligometastasis is perhaps another example: a ‘diagnosis as an alibi for treatment’ by ablation.