Christopher Bowman

Modelling strategies for controlling SARS outbreaks

Severe acute respiratory syndrome (SARS), a new, highly contagious, viral disease, emerged in China late in 2002 and quickly spread to 32 countries and regions causing in excess of 774 deaths and 8098 infections worldwide. In the absence of a rapid diagnostic test, therapy or vaccine, isolation of individuals diagnosed with SARS and quarantine of individuals feared exposed to SARS virus were used to control the spread of infection. We examine mathematically the impact of isolation and quarantine on the control of SARS during the outbreaks in Toronto, Hong Kong, Singapore and Beijing using a deterministic model that closely mimics the data for cumulative infected cases and SARS–related deaths in the first three regions but not in Beijing until mid–April, when China started to report data more accurately. The results reveal that achieving a reduction in the contact rate between susceptible and diseased individuals by isolating the latter is a critically important strategy that can control SARS outbreaks with or without quarantine. An optimal isolation programme entails timely implementation under stringent hygienic precautions defined by a critical threshold value. Values below this threshold lead to control, but those above are associated with the incidence of new community outbreaks or nosocomial infections, a known cause for the spread of SARS in each region. Allocation of resources to implement optimal isolation is more effective than to implement sub–optimal isolation and quarantine together. A community–wide eradication of SARS is feasible if optimal isolation is combined with a highly effective screening programme at the points of entry.

A Vaccination Model for Transmission Dynamics of Influenza

Despite the availability of preventive vaccines and public health vaccination programs, influenza inflicts substantial morbidity, mortality, and socio-economic costs and remains a major public heal...

Population-Wide Emergence of Antiviral Resistance during Pandemic Influenza

Background The emergence of neuraminidase inhibitor resistance has raised concerns about the prudent use of antiviral drugs in response to the next influenza pandemic. While resistant strains may initially emerge with compromised viral fitness, mutations that largely compensate for this impaired fitness can arise. Understanding the extent to which these mutations affect the spread of disease in the population can have important implications for developing pandemic plans. Methodology/Principal Findings By employing a deterministic mathematical model, we investigate possible scenarios for the emergence of population-wide resistance in the presence of antiviral drugs. The results show that if the treatment level (the fraction of clinical infections which receives treatment) is maintained constant during the course of the outbreak, there is an optimal level that minimizes the final size of the pandemic. However, aggressive treatment above the optimal level can substantially promote the spread of highly transmissible resistant mutants and increase the total number of infections. We demonstrate that resistant outbreaks can occur more readily when the spread of disease is further delayed by applying other curtailing measures, even if treatment levels are kept modest. However, by changing treatment levels over the course of the pandemic, it is possible to reduce the final size of the pandemic below the minimum achieved at the optimal constant level. This reduction can occur with low treatment levels during the early stages of the pandemic, followed by a sharp increase in drug-use before the virus becomes widely spread. Conclusions/Significance Our findings suggest that an adaptive antiviral strategy with conservative initial treatment levels, followed by a timely increase in the scale of drug-use, can minimize the final size of a pandemic while preventing large outbreaks of resistant infections.

Emergence of drug resistance: implications for antiviral control of pandemic influenza.

Given the danger of an unprecedented spread of the highly pathogenic avian influenza strain H5N1 in humans, and great challenges to the development of an effective influenza vaccine, antiviral drugs will probably play a pivotal role in combating a novel pandemic strain. A critical limitation to the use of these drugs is the evolution of highly transmissible drug-resistant viral mutants. Here, we develop a mathematical model to evaluate the potential impact of an antiviral treatment strategy on the emergence of drug resistance and containment of a pandemic. The results show that elimination of the wild-type strain depends crucially on both the early onset of treatment in indexed cases and population-level treatment. Given the probable delay of 0.5–1 day in seeking healthcare and therefore initiating therapy, the findings indicate that a single strategy of antiviral treatment will be unsuccessful at controlling the spread of disease if the reproduction number of the wild-type strain exceeds 1.4. We demonstrate the possible occurrence of a self-sustaining epidemic of resistant strain, in terms of its transmission fitness relative to the wild-type, and the reproduction number . Considering reproduction numbers estimated for the past three pandemics, the findings suggest that an uncontrollable pandemic is likely to occur if resistant viruses with relative transmission fitness above 0.4 emerge. While an antiviral strategy is crucial for containing a pandemic, its effectiveness depends critically on timely and strategic use of drugs.

Using multiple data features improved the validity of osteoporosis case ascertainment from administrative databases

OBJECTIVES The aim was to construct and validate algorithms for osteoporosis case ascertainment from administrative databases and to estimate the population prevalence of osteoporosis for these algorithms. STUDY DESIGN AND SETTING Artificial neural networks, classification trees, and logistic regression were applied to hospital, physician, and pharmacy data from Manitoba, Canada. Discriminative performance and calibration (i.e., error) were compared for algorithms defined from different sets of diagnosis, prescription drug, comorbidity, and demographic variables. Algorithms were validated against a regional bone mineral density testing program. RESULTS Discriminative performance and calibration were poorer and sensitivity was generally lower for algorithms based on diagnosis codes alone than for algorithms based on an expanded set of data features that included osteoporosis prescriptions and age. Validation measures were similar for neural networks and classification trees, but prevalence estimates were lower for the former model. CONCLUSION Multiple features of administrative data generally resulted in improved sensitivity of osteoporosis case-detection algorithm without loss of specificity. However, prevalence estimates using an expanded set of features were still slightly lower than estimates from a population-based study with primary data collection. The classification methods developed in this study can be extended to other chronic diseases for which there may be multiple markers in administrative data.

Antiviral resistance during pandemic influenza: implications for stockpiling and drug use

BackgroundThe anticipated extent of antiviral use during an influenza pandemic can have adverse consequences for the development of drug resistance and rationing of limited stockpiles. The strategic use of drugs is therefore a major public health concern in planning for effective pandemic responses.MethodsWe employed a mathematical model that includes both sensitive and resistant strains of a virus with pandemic potential, and applies antiviral drugs for treatment of clinical infections. Using estimated parameters in the published literature, the model was simulated for various sizes of stockpiles to evaluate the outcome of different antiviral strategies.ResultsWe demonstrated that the emergence of highly transmissible resistant strains has no significant impact on the use of available stockpiles if treatment is maintained at low levels or the reproduction number of the sensitive strain is sufficiently high. However, moderate to high treatment levels can result in a more rapid depletion of stockpiles, leading to run-out, by promoting wide-spread drug resistance. We applied an antiviral strategy that delays the onset of aggressive treatment for a certain amount of time after the onset of the outbreak. Our results show that if high treatment levels are enforced too early during the outbreak, a second wave of infections can potentially occur with a substantially larger magnitude. However, a timely implementation of wide-scale treatment can prevent resistance spread in the population, and minimize the final size of the pandemic.ConclusionOur results reveal that conservative treatment levels during the early stages of the outbreak, followed by a timely increase in the scale of drug-use, will offer an effective strategy to manage drug resistance in the population and avoid run-out. For a 1918-like strain, the findings suggest that pandemic plans should consider stockpiling antiviral drugs to cover at least 20% of the population.

Post-exposure prophylaxis during pandemic outbreaks

BackgroundWith the rise of the second pandemic wave of the novel influenza A (H1N1) virus in the current season in the Northern Hemisphere, pandemic plans are being carefully re-evaluated, particularly for the strategic use of antiviral drugs. The recent emergence of oseltamivir-resistant in treated H1N1 patients has raised concerns about the prudent use of neuraminidase inhibitors for both treatment of ill individuals and post-exposure prophylaxis of close contacts.MethodsWe extended an established population dynamical model of pandemic influenza with treatment to include post-exposure prophylaxis of close contacts. Using parameter estimates published in the literature, we simulated the model to evaluate the combined effect of treatment and prophylaxis in minimizing morbidity and mortality of pandemic infections in the context of transmissible drug resistance.ResultsWe demonstrated that, when transmissible resistant strains are present, post-exposure prophylaxis can promote the spread of resistance, especially when combined with aggressive treatment. For a given treatment level, there is an optimal coverage of prophylaxis that minimizes the total number of infections (final size) and this coverage decreases as a higher proportion of infected individuals are treated. We found that, when treatment is maintained at intermediate levels, limited post-exposure prophylaxis provides an optimal strategy for reducing the final size of the pandemic while minimizing the total number of deaths. We tested our results by performing a sensitivity analysis over a range of key model parameters and observed that the incidence of infection depends strongly on the transmission fitness of resistant strains.ConclusionOur findings suggest that, in the presence of transmissible drug resistance, strategies that prioritize the treatment of only ill individuals, rather than the prophylaxis of those suspected of being exposed, are most effective in reducing the morbidity and mortality of the pandemic. The impact of post-exposure prophylaxis depends critically on the treatment level and the transmissibility of resistant strains and, therefore, enhanced surveillance and clinical monitoring for resistant mutants constitutes a key component of any comprehensive plan for antiviral drug use during an influenza pandemic.

Automated analysis of gene-microarray images

cDNA micro-arrays are a relatively new technology that allow the viewing of gene expression of many genes (or other DNA fragments) simultaneously. The output of a micro-array experiment is a pair of digital images, each of which show thousands of individual spots. This paper introduces a novel, operator independent, and reproducible algorithm for determining the relative intensities of these spots. This method permits high throughput analysis of micro-array images, which is very desirable given the volume of data collected. This algorithm for automated spot location makes use of the regular structure of the images to produce an initial approximation of spot location, which is then iteratively refined. The algorithm will be tested on micro-array images produced at the Canadian Centre for Human and Animal Health, as well as on publicly available micro-array images.

Transmissibility of the 2009 H1N1 pandemic in remote and isolated Canadian communities: a modelling study

Objectives During the first wave of the 2009 influenza pH1N1, disease burden was distributed in a geographically heterogeneous fashion. It was particularly high in some remote and isolated Canadian communities when compared with urban centres. We sought to estimate the transmissibility (the basic reproduction number) of pH1N1 strain in some remote and isolated Canadian communities. Design A discrete time susceptible-exposed-infected transmission model was fit to infection curves simulated from laboratory-confirmed case counts for pH1N1 on each day. The sampling from Poisson distribution was used to estimate the basic reproduction number, R0, of pH1N1 during the spring wave for five different communities in Manitoba and Nunavut, Canada, where remote and isolated communities experienced a high incidence of infection, and high rates of hospitalisation and intensive care unit admission. Setting Remote and isolated communities in Northern Manitoba, Nunavut, and the largest urban centre (Winnipeg) in the province of Manitoba, Canada. Results Using published values of the exposed and infectious periods specific to H1N1 infection, corresponding to the average generation time of 2.78 days, we estimated a mean value of 2.26 for R0 (95% CI 1.57 to 3.75) in a community located in northern Manitoba. Estimates of R0 for other communities in Nunavut varied considerably with higher mean values of 3.91 (95% CI 3.08 to 4.87); 2.03 (95% CI 1.50 to 3.19); and 2.45 (95% CI 1.68 to 3.44). We estimated a lower mean value of 1.57 (95% CI 1.35 to 1.87) for R0 in the Winnipeg health region, as the largest urban centre in Manitoba. Conclusions Influenza pH1N1 appears to have been far more transmissible in rural and isolated Canadian communities than other large urban areas. The differential severity of the pandemic in these regions may be explained partly by differential transmissibility, and suggests the need for more nuanced, targeted or population-specific control strategies in Canada.

Evaluation of vaccination strategies during pandemic outbreaks

During pandemic influenza, several factors could significantly impact the outcome of vaccination campaigns, including the delay in pandemic vaccine availability, inadequate protective efficacy, and insufficient number of vaccines to cover the entire population. Here, we incorporate these factors into a vaccination model to investigate and compare the effectiveness of the single-dose and two-dose vaccine strategies. The results show that, if vaccination starts early enough after the onset of the outbreak, a two-dose strategy can lead to a greater reduction in the total number of infections. This, however, requires the second dose of vaccine to confer a substantially higher protection compared to that induced by the first dose. For a sufficiently long delay in start of vaccination, the single-dose strategy outperforms the two-dose vaccination program regardless of its protection efficacy. The findings suggest that the population-wide benefits of a single-dose strategy could in general be greater than the two-dose vaccination program, in particular when the second dose offers marginal increase in the protection induced by the first dose.