• over-diagnosis
• kawasaki disease

Coon et al,1 who have published on other areas of putative overdiagnosis, bring us the example of coronary artery abnormalities in Kawasaki disease in children. Across 48 participating hospitals, they noted that over some 15 years the rate of diagnosis of non-severe cardiac abnormalities had increased significantly by some twofold, yet adverse cardiac outcomes remained stable. Their contention is that this may represent an increasing rate of overdiagnosis and that if this is the case, some children may be unnecessarily exposed to potential harm. The issue of overdiagnosis in all age groups has received increased attention recently,2 3 so in this Editorial I will explore some of the issues as they relate to child health and consider the societal, cultural and professional drivers.

Overdiagnosis has no formal definition,4 5 but for the purpose of this piece, we can think of it as ‘the finding of mild disease for which the harms in diagnosing and treating exceed the benefits’. Nor is the issue a new one: Black and Welch published their landmark paper on the overestimation of disease prevalence 25 years ago, and even then they saw the problem as having crept up over the previous two decades.6 For fairly obvious reasons the issue has been of particular concern to radiologists as techniques and modalities have become progressively more refined, so more things are found on imaging that are not necessarily diseases.7

We can think of overdiagnosis as part of a spectrum of ‘Too Much Medicine’ that includes over-referral, overinvestigation, the invention of disease entities and ‘treatments’ for them and either mistaking pathology for physiology or becoming intolerant of the extremes of physiological variability. It takes place in, and partly arises from, exaggerated societal expectations of medicine together with worries about health and disease in our age of anxiety. Think of plagiocephaly and helmets, and treatments both to limit final height and ‘improve’ short stature. However, there are other drivers of overdiagnosis within medical culture. How often have we heard non-explanations for practices such as ‘This is what I was taught’ and ‘This is how we do things here’?

Overdiagnosis begins before birth and is one of the unintended consequences of fetal anomaly screening. Examples include mild but resolving cerebral ventriculomegaly, mild hydronephrosis and disappearing cystic lung lesions. The subsequent interventions they give rise to are relatively benign: respectively, head ultrasound, renal ultrasound and chest X-ray. However, the parental anxiety generated ahead of the ‘all clear’ can be significant, and some babies may end up with more invasive investigations or ones needing higher radiation doses before ‘normality’ is confirmed.

I suggest that overdiagnosis may sometimes be an unintended side effect of efforts to standardise practice through, for example, National Institute for Health and Care Excellence (NICE) guidelines. Many paediatricians have been concerned about the effect of the recent NICE guideline (CG 98) on the management of neonatal jaundice. Compared with preguideline days, babies with normal physiological jaundice can end up labelled as having ‘pathology’ for which they must be ‘treated’. In spite of the measured tone of the guideline document itself, strict application of the recommendations as currently framed seems to result in excessive resort to phototherapy, extended hospital stays and multiple heel prick bilirubin measurements in a significant number of babies who previously would have avoided much of this. While we have no idea of the number needed to treat more aggressively in order to prevent a case of kernicterus, paediatricians have an everyday awareness of those babies paying the price of close guideline adherence. In spite of this, I have found no publication that reports a good quality assessment of the guideline’s impact in the UK. Analogous concerns have been raised about CG 149: ‘Neonatal infection (early onset): antibiotics for prevention and treatment’.

One of the dangers of consensus guidelines is that they have the potential to sanctify habitual practice rather than optimise practice based on evidence. Imagine two practices that are used in a particular condition: let’s call one ‘standard’ and the other ‘variant’, and let’s say that neither has been tested against the other to assess effectiveness or harm. Moreover, for the sake of argument, let’s assume that the ‘variant’ practice will one day be found to be better, because it turns out to be less harmful though no more effective. The chances are that a consensus guideline will opt for ‘standard’ practice. The consequence will be that more children will be harmed until such time as the trial takes place, whereas if both practices were allowed to coexist, fewer children would be harmed. Which then is the more ethically appropriate stance: to retain plurality of practice or to standardise merely to reduce variation?

Even when a guideline is based on component parts that have a reasonable evidence base, there can still be a good case for subjecting the guideline as a whole to a rigorous real-world evaluation, because the evidence for individual components of a care package may not straightforwardly predict the effect on patient outcomes of the package as a whole. The case for this approach is especially strong when the evidence base for guideline creation is largely observational data. Instead, in the aftermath of guideline publication, there are more commonly publications assessing process (were the guidelines properly implemented?) rather than outcome (were patients better off as a result?).

It is easy to bring to mind examples within paediatrics where we may be doing too much. An obvious one is the use of antibiotics in viral respiratory infections, especially bronchiolitis. Other examples may sit under our noses, unnoticed because routine and unquestioned: the frequency with which routine haematology and biochemistry are measured, and giving fluid or medication intravenously when the enteral route would be just as good. However, there have been areas of progress and achievement: for example, far fewer children have their tonsils removed or grommets inserted than a generation ago.

Many of the physician behaviours that result in overdiagnosis reflect anxiety about uncertainty, anxiety about the consequences of missing something, anxiety about compliance with guidelines and anxiety about regulators and inspectors. Yet medical, and especially paediatric, practice takes place in a sea of uncertainty: diffuse or intermittent symptoms, borderline or equivocal tests, unanticipated turns of events and complicated patients and parents whose real issues may not be the ones they first tell you about.

Seen from the perspective of a trainee, one aspect of growing up, in a professional sense, consists of learning to live with uncertainty and manage anxiety so that it does not drive inappropriate medical behaviours. Yet better understanding of the mechanisms of overdiagnosis will not necessarily lead to changes in professional behaviour such that the problem would suddenly disappear. Going back to coronary aneurysms in Kawasaki disease, what could have helped the overdiagnosed children might have been some hard data to demonstrate that it would be safe not to act on small lesions that previously went unrecognised. With that knowledge, clinicians might have the confidence to ignore them.