The global transmission and control of influenza

New strains of influenza spread around the globe via the movement of infected individuals. The global dynamics of influenza are complicated by different patterns of influenza seasonality in different regions of the world. We have released an open-source stochastic mathematical model of the spread of influenza across 321 major, strategically located cities of the world. Influenza is transmitted between cities via infected airline passengers. Seasonality is simulated by increasing the transmissibility in each city at the times of the year when influenza has been observed to be most prevalent. The spatiotemporal spread of pandemic influenza can be understood through clusters of global transmission and links between them, which we identify using the epidemic percolation network (EPN) of the model. We use the model to explain the observed global pattern of spread for pandemic influenza A(H1N1) 2009-2010 (pandemic H1N1 2009) and to examine possible global patterns of spread for future pandemics depending on the origin of pandemic spread, time of year of emergence, and basic reproductive number (). We also use the model to investigate the effectiveness of a plausible global distribution of vaccine for various pandemic scenarios. For pandemic H1N1 2009, we show that the biggest impact of vaccination was in the temperate northern hemisphere. For pandemics starting in the temperate northern hemisphere in May or April, vaccination would have little effect in the temperate southern hemisphere and a small effect in the tropics. With the increasing interconnectedness of the world's population, we must take a global view of infectious disease transmission. Our open-source, computationally simple model can help public health officials plan for the next pandemic as well as deal with interpandemic influenza.     

Optimal control for pandemic influenza: The role of limited antiviral treatment and isolation

The implementation of optimal control strategies involving antiviral treatment and/or isolation measures can reduce significantly the number of clinical cases of influenza. Pandemic-level control measures must be carefully assessed specially in resource-limited situations. A model for the transmission dynamics of influenza is used to evaluate the impact of isolation and/or antiviral drug delivery measures during an influenza pandemic. Five pre-selected control strategies involving antiviral treatment and isolation are tested under the “unlimited” resource assumption followed by an exploration of the impact of these “optimal” policies when resources are limited in the context of a 1918-type influenza pandemic scenario. The implementation of antiviral treatment at the start of a pandemic tends to reduce the magnitude of epidemic peaks, spreading the maximal impact of an outbreak over an extended window in time. Hence, the controls’ timing and intensity can reduce the pressures placed on the health care infrastructure by a pandemic reducing the stress put on the system during epidemic peaks. The role of isolation strategies is highlighted in this study particularly when access to antiviral resources is limited.     

Qualitative analysis of the level of cross-protection between epidemic waves of the 1918-1919 influenza pandemic

The 1918-1919 influenza pandemic was composed of multiple waves within a period of nine months in several regions of the world. Increasing our understanding of the mechanisms responsible for this multi-wave profile has important public health implications. We model the transmission dynamics of two strains of influenza interacting via cross-immunity to simulate two temporal waves of influenza and explore the impact of the basic reproduction number, as a measure of transmissibility associated to each influenza strain, cross-immunity and the timing of the onset of the second influenza epidemic on the pandemic profile. We use time series of case notifications during the 1918 influenza pandemic in Geneva, Switzerland, for illustration. We calibrate our mathematical model to the initial wave of infection to estimate the basic reproduction number of the first wave and the corresponding timing of onset of the second influenza variant. We use this information to explore the impact of cross-immunity levels on the dynamics of the second wave of influenza. Our results for the 1918 pandemic in Geneva, Switzerland, indicate that a second wave can occur whenever R₀₁<1.5 or when cross-immunity levels are less than 0.58 for our estimated R₀₂ of 2.4. We also explore qualitatively profiles of two-wave pandemics and compare them with real temporal profiles of the 1918 influenza pandemic in other regions of the world including several Scandinavian cities, New York City, England and Wales, and Sydney, Australia. Pandemic profiles are classified into three broad categories namely “right-handed”, “left-handed”, and “M-shape”. Our results indicate that avoiding a second influenza epidemic is plausible given sufficient levels of cross-protection are attained via natural infection during an early (herald) wave of infection or vaccination campaigns prior to a second wave. Furthermore, interventions aimed at mitigating the first pandemic wave may be counterproductive by increasing the chances of a second wave of infection that could potentially be more virulent than the first.     

Pandemic (H1N1) 2009 Influenza Community Transmission Was Established in One Australian State When the Virus Was First Identified in North America

Background

In mid-June 2009 the State of Victoria in Australia appeared to have the highest notification rate of pandemic (H1N1) 2009 influenza in the world. We hypothesise that this was because community transmission of pandemic influenza was already well established in Victoria at the time testing for the novel virus commenced. In contrast, this was not true for the pandemic in other parts of Australia, including Western Australia (WA).

Methods

We used data from detailed case follow-up of patients with confirmed infection in Victoria and WA to demonstrate the difference in the pandemic curve in two Australian states on opposite sides of the continent. We modelled the pandemic in both states, using a susceptible-infected-removed model with Bayesian inference accounting for imported cases.

Results

Epidemic transmission occurred earlier in Victoria and later in WA. Only 5% of the first 100 Victorian cases were not locally acquired and three of these were brothers in one family. By contrast, 53% of the first 102 cases in WA were associated with importation from Victoria. Using plausible model input data, estimation of the effective reproductive number for the Victorian epidemic required us to invoke an earlier date for commencement of transmission to explain the observed data. This was not required in modelling the epidemic in WA.

Conclusion

Strong circumstantial evidence, supported by modelling, suggests community transmission of pandemic influenza was well established in Victoria, but not in WA, at the time testing for the novel virus commenced in Australia. The virus is likely to have entered Victoria and already become established around the time it was first identified in the US and Mexico.     

Influenza outbreaks

Influenza is a virus that causes considerable morbidity and mortality in human populations every year. This fact, coupled with its perceived pandemic potential, means that influenza features prominently in both scientific literature and the media. In this review we focus on the biological assumptions behind theoretical attempts to understand the seasonal and evolutionary dynamics of influenza through mathematical modelling and suggest that the largely unchallenged dogma upon which most efforts are currently based is sorely lacking.     

Can Interactions between Timing of Vaccine-Altered Influenza Pandemic Waves and Seasonality in Influenza Complications Lead to More Severe Outcomes?

Vaccination can delay the peak of a pandemic influenza wave by reducing the number of individuals initially susceptible to influenza infection. Emerging evidence indicates that susceptibility to severe secondary bacterial infections following a primary influenza infection may vary seasonally, with peak susceptibility occurring in winter. Taken together, these two observations suggest that vaccinating to prevent a fall pandemic wave might delay it long enough to inadvertently increase influenza infections in winter, when primary influenza infection is more likely to cause severe outcomes. This could potentially cause a net increase in severe outcomes. Most pandemic models implicitly assume that the probability of severe outcomes does not vary seasonally and hence cannot capture this effect. Here we show that the probability of intensive care unit (ICU) admission per influenza infection in the 2009 H1N1 pandemic followed a seasonal pattern. We combine this with an influenza transmission model to investigate conditions under which a vaccination program could inadvertently shift influenza susceptibility to months where the risk of ICU admission due to influenza is higher. We find that vaccination in advance of a fall pandemic wave can actually increase the number of ICU admissions in situations where antigenic drift is sufficiently rapid or where importation of a cross-reactive strain is possible. Moreover, this effect is stronger for vaccination programs that prevent more primary influenza infections. Sensitivity analysis indicates several mechanisms that may cause this effect. We also find that the predicted number of ICU admissions changes dramatically depending on whether the probability of ICU admission varies seasonally, or whether it is held constant. These results suggest that pandemic planning should explore the potential interactions between seasonally varying susceptibility to severe influenza outcomes and the timing of vaccine-altered pandemic influenza waves.     

FluTE,a Publicly Available Stochastic Influenza Epidemic Simulation Model

Mathematical and computer models of epidemics have contributed to our understanding of the spread of infectious disease and the measures needed to contain or mitigate them. To help prepare for future influenza seasonal epidemics or pandemics, we developed a new stochastic model of the spread of influenza across a large population. Individuals in this model have realistic social contact networks, and transmission and infections are based on the current state of knowledge of the natural history of influenza. The model has been calibrated so that outcomes are consistent with the 1957/1958 Asian A(H2N2) and 2009 pandemic A(H1N1) influenza viruses. We present examples of how this model can be used to study the dynamics of influenza epidemics in the United States and simulate how to mitigate or delay them using pharmaceutical interventions and social distancing measures. Computer simulation models play an essential role in informing public policy and evaluating pandemic preparedness plans. We have made the source code of this model publicly available to encourage its use and further development.     

Comparing pandemic to seasonal influenza mortality: moderate impact overall but high mortality in young children

Background

We assessed the severity of the 2009 influenza pandemic by comparing pandemic mortality to seasonal influenza mortality. However, reported pandemic deaths were laboratory-confirmed – and thus an underestimation – whereas seasonal influenza mortality is often more inclusively estimated. For a valid comparison, our study used the same statistical methodology and data types to estimate pandemic and seasonal influenza mortality.

Methods and Findings

We used data on all-cause mortality (1999–2010, 100% coverage, 16.5 million Dutch population) and influenza-like-illness (ILI) incidence (0.8% coverage). Data was aggregated by week and age category. Using generalized estimating equation regression models, we attributed mortality to influenza by associating mortality with ILI-incidence, while adjusting for annual shifts in association. We also adjusted for respiratory syncytial virus, hot/cold weather, other seasonal factors and autocorrelation. For the 2009 pandemic season, we estimated 612 (range 266–958) influenza-attributed deaths; for seasonal influenza 1,956 (range 0–3,990). 15,845 years-of-life-lost were estimated for the pandemic; for an average seasonal epidemic 17,908. For 0–4 yrs of age the number of influenza-attributed deaths during the pandemic were higher than in any seasonal epidemic; 77 deaths (range 61–93) compared to 16 deaths (range 0–45). The ≥75 yrs of age showed a far below average number of deaths. Using pneumonia/influenza and respiratory/cardiovascular instead of all-cause deaths consistently resulted in relatively low total pandemic mortality, combined with high impact in the youngest age category.

Conclusion

The pandemic had an overall moderate impact on mortality compared to 10 preceding seasonal epidemics, with higher mortality in young children and low mortality in the elderly. This resulted in a total number of pandemic deaths far below the average for seasonal influenza, and a total number of years-of-life-lost somewhat below average. Comparing pandemic and seasonal influenza mortality as in our study will help assessing the worldwide impact of the 2009 pandemic.     

Pandemic Dynamics and the Breakdown of Herd Immunity

In this note we discuss the issues involved in attempting to model pandemic dynamics. More specifically, we show how it may be possible to make projections for the ongoing H1N1 pandemic as extrapolated from knowledge of seasonal influenza. We derive first-approximation parameter estimates for the SIR model to describe seasonal influenza, and then explore the implications of the existing classical epidemiological theory for the case of a pandemic virus. In particular, we note the dramatic nonlinear increase in attack rate as a function of the percentage of susceptibles initially present in the population. This has severe consequences for the pandemic, given the general lack of immunity in the global population.     

Pandemic influenza: overview of vaccines and antiviral drugs

Pandemic influenza has become a high priority item for all public health authorities. An influenza pandemic is believed to be imminent, and scientists agree that it will be a matter of when, where, and what will be the causative agent. Recently, most attention has been directed to human cases of avian influenza caused by a H5N1 avian influenza virus. An effective vaccine will be needed to substantially reduce the impact of an influenza pandemic. Current influenza vaccine manufacturing technology is not adequate to support vaccine production in the event of an avian influenza outbreak, and it has now become clear that new innovative production technology is required. Antiviral drugs, on the other hand, can play a very important role in slowing the disease spread but are in short supply and resistance has been a major issue. Here, we provide an update on the status of pandemic vaccine development and antiviral drugs. Finally, we conclude with some proposed areas of focus in pandemic vaccine preparedness.     

The effect of risk perception on the 2009 H1N1 pandemic influenza dynamics

Background

The 2009 H1N1 pandemic influenza dynamics in Italy was characterized by a notable pattern: as it emerged from the analysis of influenza-like illness data, after an initial period (September–mid-October 2009) characterized by a slow exponential increase in the weekly incidence, a sudden and sharp increase of the growth rate was observed by mid-October. The aim here is to understand whether spontaneous behavioral changes in the population could be responsible for such a pattern of epidemic spread.

Methodology/Principal Findings

In order to face this issue, a mathematical model of influenza transmission, accounting for spontaneous behavioral changes driven by cost/benefit considerations on the perceived risk of infection, is proposed and validated against empirical epidemiological data. The performed investigation revealed that an initial overestimation of the risk of infection in the general population, possibly induced by the high concern for the emergence of a new influenza pandemic, results in a pattern of spread compliant with the observed one. This finding is also supported by the analysis of antiviral drugs purchase over the epidemic period. Moreover, by assuming a generation time of 2.5 days, the initially diffuse misperception of the risk of infection led to a relatively low value of the reproductive number , which increased to in the subsequent phase of the pandemic.

Conclusions/Significance

This study highlights that spontaneous behavioral changes in the population, not accounted by the large majority of influenza transmission models, can not be neglected to correctly inform public health decisions. In fact, individual choices can drastically affect the epidemic spread, by altering timing, dynamics and overall number of cases.     

Using a Dynamic Model to Consider Optimal Antiviral Stockpile Size in the Face of Pandemic Influenza Uncertainty

Background

The Canadian National Antiviral Stockpile (NAS) contains treatment for 17.5% of Canadians. This assumes no concurrent intervention strategies and no wastage due to non-influenza respiratory infections. A dynamic model can provide a mechanism to consider complex scenarios to support decisions regarding the optimal NAS size under uncertainty.

Methods

We developed a dynamic model for pandemic influenza in Canada that is structured by age and risk to calculate the demand for antivirals to treat persons with pandemic influenza under a wide-range of scenarios that incorporated transmission dynamics, disease severity, and intervention strategies. The anticipated per capita number of acute respiratory infections due to viruses other than influenza was estimated for the full pandemic period from surveys based on criteria to identify potential respiratory infections.

Results

Our results demonstrate that up to two thirds of the population could develop respiratory symptoms as a result of infection with a pandemic strain. In the case of perfect antiviral allocation, up to 39.8% of the population could request antiviral treatment. As transmission dynamics, severity and timing of the emergence of a novel influenza strain are unknown, the sensitivity analysis produced considerable variation in potential demand (median: 11%, IQR: 2–21%). If the next pandemic strain emerges in late spring or summer and a vaccine is available before the anticipated fall wave, the median prediction was reduced to 6% and IQR to 0.7–14%. Under the strategy of offering empirical treatment to all patients with influenza like symptoms who present for care, demand could increase to between 65 and 144%.

Conclusions

The demand for antivirals during a pandemic is uncertain. Unless an accurate, timely and cost-effective test is available to identify influenza cases, demand for antivirals from persons infected with other respiratory viruses will be substantial and have a significant impact on the NAS.     

H5N1 and 1918 pandemic influenza virus infection results in early and excessive infiltration of macrophages and neutrophils in the lungs of mice

Fatal human respiratory disease associated with the 1918 pandemic influenza virus and potentially pandemic H5N1 viruses is characterized by severe lung pathology, including pulmonary edema and extensive inflammatory infiltrate. Here, we quantified the cellular immune response to infection in the mouse lung by flow cytometry and demonstrate that mice infected with highly pathogenic (HP) H1N1 and H5N1 influenza viruses exhibit significantly high numbers of macrophages and neutrophils in the lungs compared to mice infected with low pathogenic (LP) viruses. Mice infected with the 1918 pandemic virus and a recent H5N1 human isolate show considerable similarities in overall lung cellularity, lung immune cell sub-population composition, and cellular immune temporal dynamics. Interestingly, while these similarities were observed, the HP H5N1 virus consistently elicited significantly higher levels of pro-inflammatory cytokines in whole lungs and primary human macrophages, revealing a potentially critical difference in the pathogenesis of H5N1 infections. Primary mouse and human macrophages and dendritic cells were also susceptible to 1918 and H5N1 influenza virus infection in vitro. These results together indicate that infection with HP influenza viruses such as H5N1 and the 1918 pandemic virus leads to a rapid cell recruitment of macrophages and neutrophils into the lungs, suggesting that these cells play a role in acute lung inflammation associated with HP influenza virus infection.     

An evolutionary model of influenza A with drift and shift

Influenza A virus evolves through two types of evolutionary mechanisms - drift and shift. These two evolutionary mechanisms allow the pathogen to infect us repeatedly, as well as occasionally create pandemics with large morbidity and mortality. Here we introduce a novel model that incorporates both evolutionary mechanisms. This necessitates the modelling of three types of strains - seasonal human strains, bird-to-human transmittable H5N1 strains and evolved pandemic H5N1 strain. We define reproduction and invasion reproduction numbers and use them to establish the presence of dominant and coexistence equilibria. We find that the amino acid substitution structure of human influenza can destabilize the human influenza equilibrium and sustained oscillations are possible. We find that for low levels of infection in domestic birds, these oscillations persist, inducing oscillations in the number of humans infected with the avian flu strain. The oscillations have a period of 365 days, similar to the one that can be observed in the cumulative number of human H5N1 cases reported by the World Health Organization (WHO). Furthermore, we establish some partial global results on the competition of the strains.     

The Effects of School Closures on Influenza Outbreaks and Pandemics: Systematic Review of Simulation Studies

Background

School closure is a potential intervention during an influenza pandemic and has been investigated in many modelling studies.

Objectives

To systematically review the effects of school closure on influenza outbreaks as predicted by simulation studies.

Methods

We searched Medline and Embase for relevant modelling studies published by the end of October 2012, and handsearched key journals. We summarised the predicted effects of school closure on the peak and cumulative attack rates and the duration of the epidemic. We investigated how these predictions depended on the basic reproduction number, the timing and duration of closure and the assumed effects of school closures on contact patterns.

Results

School closures were usually predicted to be most effective if they caused large reductions in contact, if transmissibility was low (e.g. a basic reproduction number <2), and if attack rates were higher in children than in adults. The cumulative attack rate was expected to change less than the peak, but quantitative predictions varied (e.g. reductions in the peak were frequently 20–60% but some studies predicted >90% reductions or even increases under certain assumptions). This partly reflected differences in model assumptions, such as those regarding population contact patterns.

Conclusions

Simulation studies suggest that school closure can be a useful control measure during an influenza pandemic, particularly for reducing peak demand on health services. However, it is difficult to accurately quantify the likely benefits. Further studies of the effects of reactive school closures on contact patterns are needed to improve the accuracy of model predictions.     

Protection from lethal influenza virus challenge by oral type 1 interferon

The persistence of highly pathogenic avian influenza within wild bird populations has forged interest in control measures to limit a possible human pandemic. We therefore investigated the efficacy of low dose oral administration of IFN-alpha as a potential therapy against influenza infection in a murine model. We have identified an optimal low oral dose of IFN-alpha that when delivered daily as prophylactic therapy protects C57BL/6J mice from a lethal challenge with mouse adapted human influenza virus A/PR/8/34 (H1N1). These results provide strong support for the application of low dose type 1 IFN pretreatment to human influenza control.     

In Silico Structural Homology Modelling and Docking for Assessment of Pandemic Potential of a Novel H7N9 Influenza Virus and Its Ability to Be Neutralized by Existing Anti-Hemagglutinin Antibodies

The unpredictable nature of pandemic influenza and difficulties in early prediction of pandemic potential of new isolates present a major challenge for health planners. Vaccine manufacturers, in particular, are reluctant to commit resources to development of a new vaccine until after a pandemic is declared. We hypothesized that a structural bioinformatics approach utilising homology-based molecular modelling and docking approaches would assist prediction of pandemic potential of new influenza strains alongside more traditional laboratory and sequence-based methods. The newly emerged Chinese A/Hangzhou/1/2013 (H7N9) influenza virus provided a real-life opportunity to test this hypothesis. We used sequence data and a homology-based approach to construct a 3D-structural model of H7-Hangzhou hemagglutinin (HA) protein. This model was then used to perform docking to human and avian sialic acid receptors to assess respective binding affinities. The model was also used to perform docking simulations with known neutralizing antibodies to assess their ability to neutralize the newly emerged virus. The model predicted H7N9 could bind to human sialic acid receptors thereby indicating pandemic potential. The model also confirmed that existing antibodies against the HA head region are unable to neutralise H7N9 whereas antibodies, e.g. Cr9114, targeting the HA stalk region should bind with high affinity to H7N9. This indicates that existing stalk antibodies initially raised against H5N1 or other influenza A viruses could be therapeutically beneficial in prevention and/or treatment of H7N9 infections. The subsequent publication of the H7N9 HA crystal structure confirmed the accuracy of our in-silico structural model. Antibody docking studies performed using the H7N9 HA crystal structure supported the model''s prediction that existing stalk antibodies could cross-neutralise the H7N9 virus. This study demonstrates the value of using in-silico structural modelling approaches to complement physical studies in characterization of new influenza viruses.     

Patterns of spread of influenza A in Canada

Understanding spatial patterns of influenza transmission is important for designing control measures. We investigate spatial patterns of laboratory-confirmed influenza A across Canada from October 1999 to August 2012. A statistical analysis (generalized linear model) of the seasonal epidemics in this time period establishes a clear spatio-temporal pattern, with influenza emerging earlier in western provinces. Early emergence is also correlated with low temperature and low absolute humidity in the autumn. For the richer data from the 2009 pandemic, a mechanistic mathematical analysis, based on a transmission model, shows that both school terms and weather had important effects on pandemic influenza transmission.     

Transmission of a 2009 H1N1 Pandemic Influenza Virus Occurs before Fever Is Detected, in the Ferret Model

During the early phase of the 2009 influenza pandemic, attempts were made to contain the spread of the virus. Success of reactive control measures may be compromised if the proportion of transmission that occurs before overt clinical symptoms develop is high. In this study we investigated the timing of transmission of an early prototypic strain of pandemic H1N1 2009 influenza virus in the ferret model. Ferrets are the only animal model in which this can be assessed because they display typical influenza-like clinical signs including fever and sneezing after infection. We assessed transmission from infected animals to sentinels that were placed either in direct contact or in adjacent cages, the latter reflecting the respiratory droplet (RD) transmission route. We found that pre-symptomatic influenza transmission occurred via both contact and respiratory droplet exposure before the earliest clinical sign, fever, developed. Three of 3 animals exposed in direct contact between day 1 and 2 after infection of the donor animals became infected, and 2/3 of the animals exposed at this time period by the RD route acquired the infection, with the third animal becoming seropositive indicating either a low level infection or significant exposure. Moreover, this efficient transmission did not temporally correlate with respiratory symptoms, such as coughs and sneezes, but rather with the peak viral titre in the nose. Indeed respiratory droplet transmission did not occur late in infection, even though this was when sneezing and coughing were most apparent. None of the 3 animals exposed at this time by the RD route became infected and these animals remained seronegative at the end of the experiment. These data have important implications for pandemic planning strategies and suggest that successful containment is highly unlikely for a human-adapted influenza virus that transmits efficiently within a population.