Research into the dynamics of human infection with influenza viruses has been spurred on in recent years by outbreaks of avian influenza (also known as bird flu) that have led to human fatalities. However studies into these avian viruses, such as H5N1, have led to some controversy. In 2012, two studies were published on genetic mutations that enable easy transmission of H5N1 between ferrets – a mammalian model commonly used to study influenza infection due to similarities to human disease pathology. A debate was sparked over the safety of conducting such research, with concern over the potential escape of a mutant virus triggering a human pandemic. In order to carry out research into infectious agents like influenza, Biological Safety Level (BSL) laboratories have been constructed to contain them. However the possibility of a breach in containment is always present. Alessandro Vespignani from Northeastern University, USA, Stefano Merler from the Bruno Kessler Foundation, Italy, and colleagues take a computational approach to address how likely it is that the escape of a transmissible virus would not be contained within the local community, as published in a recent study in BMC Medicine. Vespignani and Merler explain more about how they constructed their model, what it revealed and what impact this may have on research into influenza viruses.
In your opinion, how likely is the accidental escape of a dangerous virus from a lab?
There are estimates about the probability of accidental escape from biosafety level (BSL) facilities that place such an event in the low probability bracket – 0.3 percent risk of release per lab per year (Klotz LC & Sylvester EJ, Bulletin of the Atomic Scientists, 2012, Aug 7). However if we consider the increasing number of BSL laboratories the combined probability to have at least one event in a ten-year window is appreciable. This probability has then to be weighted against the potential destructive effects that such an event could have on the population and the risk that the escape event could escalate to full-scale pandemic proportions. Another element of the multifaceted risk evaluation of such an event is indeed the likelihood of containment of an escape event.
What do your findings tell us about the chances of containing a virus once it has escaped from the lab?
We find from our computational model that the likelihood of containment depends on both the specific transmissibility of the pathogen and the timely implementation of non-pharmaceutical interventions such as isolation of cases, household quarantine of secondary cases and reactive/proactive school closure. The timely implementation of these interventions would result in a containment probability of greater than 90 percent for pathogens with moderate transmissibility. For more highly transmissible pathogens the containment probability decreases rapidly and is below 50 percent for pathogens with a transmissibility comparable to that of the 1918 pandemic.
Could you explain how your mathematical model was used to predict the spread of an escaped virus, and summarize your key findings?
We have been working with a very detailed, spatially explicit, stochastic individual-based model. The model integrates highly detailed data on country-specific sociodemographic structures such as household size and composition, age structure, rates of school attendance etc. We also explicitly considered that the epidemic originates in a BSL laboratory and that the initial infection of the first infected laboratory worker may generate an initial warning, resulting in a set of medical/epidemiological analyses being conducted very early on to identify the origin of the reported symptoms. The model is able to simulate the spread of the pathogen at an individual level and allows the computational modeling of containment policies. The model also considers the stochasticity of the infection dynamic, providing for each point in space and time an ensemble of possible epidemic evolutions. The statistical analysis of the possible epidemic evolutions yields an estimate of the outbreak containment probability according to different scenarios.
Our main findings suggest that the controllability of escape events is not guaranteed across a wide range of pathogen transmissibility, even in the case of optimal implementation of the non-pharmaceutical interventions. We also observe large variations in the containment probability as a function of the sociodemographic structure of the population where the outbreak starts. Generally, rural areas allow a fivefold increase in the controllability of the event if the containment policies can be implemented as effectively as in urban areas.
Your study focused on the escape of a novel influenza virus. Do you think a similar model could be applied to other types of viruses?
Influenza has a relatively short generation time and a large fraction of non-detectable transmissions. It therefore represents a major risk for pandemic escalation. For this reason we focused our study on potential pandemic influenza viruses by explicitly considering an individual-based model of influenza transmission. However the agent-based modeling framework presented can also be used for other types of viruses. In order to do that it is crucial to have a detailed knowledge of the specific transmission mechanism and the natural history of the disease that must be accurately plugged into the model.
In light of your results, what preparations should be made to best tackle the laboratory escape of a virus?
What emerges from the study is that immediate set up of all the containment measures is extremely important. Although some of the non-pharmaceutical policies also have a disruptive impact on society (school and workplace closures, household quarantines), their implementation should be considered right away. Furthermore, in our study we did not consider pharmaceutical interventions such as the preventive immunization of laboratory workers or the massive use of antiviral drugs. Although those measures may help in considerably increasing the containment probability, the availability and effectiveness of a vaccine against laboratory manipulated influenza viruses is not clear, nor is the logistical feasibility of massive antiviral interventions. For other types of viruses however, containment measures such as preventive and reactive vaccination or the use of specific pharmaceutical prophylaxis may be available and should therefore be included in the model.
How do you think your research and similar studies will impact policy decisions on influenza virus research, and specifically on the building of labs carrying out this research?
Our work provides numerical evaluation of one containment probability taking into account some of the many elements that contribute to the risk analysis associated with work carried out in BSL laboratories. However, there are many elements that have to be considered in formulating policies regarding BSL laboratories, such as population vulnerability to infectious agents, structure of the health system, and the possibility of putting in place a rapid response program. It is also worth remembering that in many cases the experiments or work carried out in those laboratories are meant to augment our capabilities for fighting these infectious diseases. We hope as ‘number providers’ that our work will provide valuable information to policy makers engaged in such a complex and multidisciplinary decision making process.