Helen Macdonald discusses the key points from the 2nd day of Preventing Overdiagnosis
Fixing overdiagnosis goes to the top. How can authorities or regulators help? Paul Glasziou (Professor and GP) proposed several (fairly simple sounding) solutions. 1) Be sure that you have sufficient understanding of the consequences (benefits and harms), before starting something new such as a screening programme. 2) Evidence helps. If there is uncertainty consider further research, including within a programme. 3) People in charge should be prepared to change their mind. Decisions, particularly those made by small groups, can turn out to be wrong. He explained that Japanese authorities stopped a screening programme for neuroblastoma in infants, when new evidence made it clear that it was not worthwhile.
Those in charge should also be careful when expanding disease definitions via guidelines. This can happen if a definition is widened to include pre-disease, lower thresholds for disease, or earlier or new diagnostic methods are used. Those in charge of guidelines need to be sure that the definition of a condition is clear. If it is changed, understand how many people will be affected. Consider why the change is happening, and what the balance of benefits and harms of a new definition are.
Henrik Vogt explained how “Big data” can promise to tell us everything about ourselves, but may succeed in telling us little. “Big data” is somewhat mysterious. Often it means data that give a lot of information on several variables, from a fancy new source such as genomics (or other -omics), detailed imaging, body functions like the content of your sweat feeding onto your smartphone, or even environmental measures.
Vogt, and statistician Claus Ekstrøm (who also talked on this topic) agreed that big data (or just data) can find abnormalities or variants in healthy people, but we often lack understanding to tell us what it means. Correlations between abnormalities (or variants) and disease, do not establish causation. There is a lot of noise, and it is hard to establish the signals. The problem is that predicting the outcome in complex biological systems (aka people) is hard.
The human genome is perhaps a case in point, said Ekstrøm. It hasn’t been able to give that much information to patients and healthcare yet—some 20 years on from its sequencing. A study screened the whole-genome sequence and used advanced imaging to look for problems in healthy adults. The study found that 78% of healthy people had a disease or risk factor. It is also possible to explore factors associated with health. But whole genome sequencing of those aged over 110 years found no specific genes associated with longevity.
Biological systems, such as humans, are complicated. Patience and time are needed, if big data is to help, and not harm. To make good decisions policy makers need to understand what happened to people who tested positive—did they get the disease or not? They also need to understand what happened to people who did not have the abnormality; did they get the disease or not?
Helen Macdonald, clinical editor, The BMJ, and general practitioner.
Competing interests: None declared.