Urs Frey, a paediatrician at the University Hospital of Berne in Switzerland, has long been interested in respiratory diseases. In particular, his observations of children with asthma had led him to speculate about the nature of the disease. “Asthma is not just a steady state, it is a dynamic system,” says Frey.

On page 667 of this issue, Frey's interest is resolved into research that could help predict both the timing and the severity of asthma attacks.

One of the keys to this work was Frey's collaboration with Béla Suki, a physicist at Boston University. Suki analyses complex nonlinear systems, such as the factors that contribute to avalanches, and wanted to apply his work to biomedical problems. Frey suggested that asthma would be a good place to start. “It is very difficult to understand why little triggers can launch big responses in asthma,” Frey says.

Frey suspected that lungs have a ‘memory’ of previous asthma attacks, which makes asthmatics more susceptible to secondary attacks. So he and Suki analysed data taken over a short period of time that measured the respiratory symptoms of a small group of infants. “Béla began to use his mathematics and my clinical questions,” Frey says. “That was an extremely fruitful collaboration.”

The results were encouraging and provided some support for Frey's hypothesis, although the data set was too small and covered too short a time span to be conclusive. Frey needed another set of results to analyse, but he didn't know where he could find what he wanted.

By chance, a conversation with Mike Silverman, a paediatrician at the University of Leicester, UK, led him to the ideal data set. About ten years ago, a team in New Zealand had examined the lung function of 80 asthmatics over three periods of six months, measuring their lung function twice a day. This was perfect for Frey and Suki, as they could analyse fluctuations in the lung function of individuals to see how these variations related to the onset and severity of asthma attacks.

What they saw was a form of long-range correlation between current lung function and the state of lung function days, weeks and even a month previously. Based on this, Suki managed to create an algorithm to calculate the risk of the next attack — an achievement that Frey describes as “a weather forecast for asthmatics”.

But the results go beyond prediction; they also shed some light on how the disease is treated. Short-acting bronchodilators are used to give quick relief from the symptoms of asthma. In some cases, they are also prescribed to be used four times a day, but are not used during the night. Frey and Suki found that this regular treatment affects the internal regulation of lung function, making the asthma less stable and harder to predict.

Frey suspects that other chronic diseases could benefit from similar algorithms — blood sugar levels could be tracked in diabetics, for instance. “If you know the internal dynamics of a chronic disease system you can use mathematical concepts from game theory to predict attacks,” says Frey.

On top of that, Frey believes this fresh view of disease could have broader implications. “Considering a chronic disease as a dynamic process could also dramatically change how drug efficacy studies are done in the future,” he says.