Risks and benefits of population screening

The concept of screening is bandied around quite a bit these days, particularly associated with genetics and genome profiling for both common and rare diseases. But the term is often used without due consideration or understanding of the issues associated with screening itself. If we are to critically assess the potential for personal genomics to prevent disease in individuals and populations, we need to understand some underlying public health principles.

So what is screening? Most importantly, medical screening generally involves testing (apparently) healthy people, rather than individuals with a particular clinical problem. This simple statement comes with quite a bit of baggage – if you go around telling healthy people that they have (or are likely to develop) some hideously awful disease, and need some rather unpleasant treatments to stop it, you really want to be sure that you’re right! Which means being certain that the test used is strongly predictive of the disease in question, rather than just being mildly associated with it, and that it can be used to reliably discriminate between those who have or will get the disease and those who don’t and won’t.

Unlike many clinical tests, screening tests often only put people into a higher (or lower) risk category, rather than offering a diagnosis. There are several reasons for this: first, the test itself may not be good enough to establish a definitive diagnosis. Because the test is offered to healthy people, is has to be as safe as possible and so is not always the most accurate test available – for example, a tissue biopsy may accurately detect cancer, but who would volunteer for one without any reason to suspect they actually have a tumour? Second, if you take someone who doesn’t have any symptoms for a disease, they are actually quite unlikely to have the disease; therefore, unless the test is 100% accurate – which none of them are – even a positive result might mean that they are still more likely not to have the disease. So a follow-on diagnostic test is usually needed. Third, in cases where screening relates to prediction of future disease, an inherent uncertainty about the future and the lack of biological determinism means that the result can only ever be probabilistic.

We all know why screening is good – it’s common sense that prediction and prevention has to be better than diagnosis and treatment. And in many cases, early treatment saves lives. But screening isn’t all good. There are myriad harms that can result from screening, particularly to the majority of people who don’t have and won’t ever suffer from the disease in question, some of whom will nonetheless be screened as positive. This can lead to follow-on tests and potentially unnecessary treatments, which can cause lasting physical and psychological harm. Interestingly, the ‘predict and prevent’ paradigm leads to the related idea that the earlier you catch it, the better off you are likely to be – which isn’t always true. Sometimes, it won’t make any difference, either because there’s nothing you can do anyway, or the disease detected would never become life-threatening, or the treatment is similarly effective regardless of when the diagnosis is made. In such cases, the main effect of screening would just be to make people live with disease for longer, rather than actually living longer.

So, what does all this have to do with genetic screening? Whether it involves testing for rare single gene disorders, or using multigenic risk prediction models for common complex diseases, the key point to remember is that screening isn’t automatically a good thing. While health systems may sometimes appear to be a bit slow on the uptake, organised population screening programmes are designed to ensure that the benefits of screening outweigh the harms. Importantly, before a screening programme can be instigated, the robust evidence is needed relating to the incidence of disease in the target population, the accuracy of the screening and diagnostic tests, and efficacy of the treatments available. Bodies such as the NHS National Screening Committee in the UK follow a specific set of criteria – initially described by Wilson and Junger in 1968 in a publication for the World Health Organisation – in order to judge whether they should offer a particular programme, and to whom. Important questions are asked, such as: Is the disease an important health problem? Does early detection actually improve prognosis? Does the treatment work? Is the programme affordable? While these questions might seem obvious, the answers aren’t, and evidence is needed to prove that the screening process offers an overall improvement in health.

A screening programme is much more than a test, and entails enormous logistical planning and organisation to ensure that all the healthcare services needed to actually prevent the disease are available to those need it, from follow-on testing and diagnosis to counselling, treatment and monitoring. A good example is the breast cancer screening programme, which is targeted at women of a certain age who are above specific absolute risk threshold (a 10-year risk of 2.3% in the UK), in whom breast cancer can be treated more successfully if detected early. Another example is the newborn blood spot screening programme, which exists in numerous countries across the globe, to look for rare but devastating inborn errors in metabolism – including phenylketonuria (PKU), the archetypal autosomal recessive genetic disease – which can be entirely prevented through dietary restriction.

Can we use genetic screening to improve population health? Definitely. And it’s already happening for certain specific diseases. Can we use personal genomes for screening to improve individual health? Maybe. While some individuals might be able to use information about personal genetic risk to modify their lifestyles, and thus prevent disease, there is very limited evidence that this is an effective long-term strategy. Before any genomic screening strategy can be considered for a public health initiative, robust evidence is needed of the direct clinical benefits, as well as the harms and costs. Ultimately, only time will tell whether genomic information will be useful for screening.