Announcements - Pandemics and probabilities

March 2, 2009 — Winnipeg, Manitoba

In 1918, vast armies returning on crowded troop ships from First World War battlefields unwittingly spread a deadly influenza virus like wildfire. Eventually, this "Spanish Flu" would kill 50 million people around the world. In the current environment of global travel, a pandemic today could be bigger, spread faster and kill more people. Historical trends suggest that the world is overdue for such a pandemic.

One statistic from 1918 is particularly intriguing: 500 million people got sick and 450 million recovered. Nobody knows why 10 percent of those affected succumbed to the disease, but Dr. Murray Alexander of the NRC Institute for Biodiagnostics (NRC-IBD) in Winnipeg says the statistical immunology and epidemiology models that his team is working on may shed light on that question, and help lower mortality rates in future epidemics.

NRC-IBD's epidemiology modelling team (left to right): Randy Summers, Dr. Murray Alexander, Dr. Christopher Bowman, Dr. Nick Pizzi, Dr. Seyed Moghadas.

It is impossible to foresee with certainty the course of a viral epidemic like the 2003 Severe Acute Respiratory Syndrome (SARS) outbreak that, in Canada, affected Vancouver and Toronto quite differently. Much like predicting weather, a startlingly large number of variables, known and unknown, affect what eventually happens.

Since its founding in 2003, the NRC research group has been designing mathematical models of virus outbreaks. Properly used, these could help public medical authorities make informed policy and operational choices that save lives.

The NRC team is investigating two ways in which viruses spread. The first considers how a virus invades the cells in a human host and subsequently evolves to escape the protective response of the immune system. The second models the social behaviours and interactions of people during outbreaks. For instance, a virus' spread may be slowed by increased hand-washing, temporary changes in social patterns or quarantines.

Currently, the NRC research team is designing separate mathematical models for both aspects of the spread of viruses. Eventually, the researchers will attempt to combine the two models. This would enable them to determine, for example, how viral-immune processes (which govern the disease course in individual hosts) ultimately affect social behaviour and, therefore, the contact patterns that spread infections within the population. In the case of measles outbreaks, the researchers have already found that incorporating interactions between the two model types sometimes leads to dramatically different disease dynamics, and the same is probably true for pandemic influenza.

"Using modelling to investigate different scenarios provides medical authorities with a 'big picture' that helps them to compare possible outcomes and chains of events, and to assess the effects of different countermeasures such as vaccination, drug intervention and quarantine," says Dr. Alexander. "Various models have consistently shown that an early response to an outbreak is crucial to increasing the chance of success in controlling — or at least mitigating — its spread in a population. Models can also help public health decision makers when faced with confounding issues," he adds. For example, if drug supplies are limited, they can determine whether more lives would be saved by giving the drug to specific groups (such as healthcare workers or the elderly), or by handing them out on a first come, first served basis.

Because the viral-immune processes and population dynamics are so intertwined, mathematicians and statisticians must view them in the context of a complex big picture, notes Dr. Alexander, and develop effective "tools" for managing this complexity and allowing for incomplete data.

So far, the NRC team has proved its models work. The push now is to expand the models to handle more data to make them more reliable predictors, and get them into the hands of health authorities. "Our expectation," says Dr. Alexander, "is that with more time to incorporate more realistic data and disease processes, these models can become even more useful."