Mathematical modelling of the pandemic H1N1 2009 (WHO WER, August 21, 2009, edited)
[Original Full Document: LINK. EDITED.]
Weekly epidemiological record - 21 august 2009, 84rd year - No. 34, 2009, 84, 341?352 - http://www.who.int/wer
Mathematical modelling of the pandemic H1N1 2009
Background and objectives of an informal network coordinated by WHO
Mathematical and computational models are useful tools for elucidating the non-linear transmission dynamics of epidemics, for analysing epidemiological data and for examining the potential impact of possible intervention options. Real-time mathematical modelling from an informal network of academic mathematical modellers coordinated by WHO during the severe acute respiratory syndrome (SARS) epidemic in 2003 yielded valuable insights, which informed strategy and control measures.
Soon after WHO declared that the emergence of a novel influenza A (H1N1) strain in Mexico and the United States represented a ?public health emergency of international concern?, WHO convened another informal network of mathematical modellers with the following goals:
(a) to describe and predict the behaviour and impact of the pandemic H1N1 2009 and demonstrate the potential outcome of proposed pharmaceutical and non-pharmaceutical interventions in different settings;
(b) to present these analyses in formats suitable for various audiences, including technical experts, policy-makers and the general public; and
(c) to adapt models and interpret experiences from developed countries so they can be applied in low-resource countries.
The mathematical modelling network met in Geneva in early July 2009. The network consists of representatives from infectious disease modelling groups (academic and public health institutions), professionals from national public health agencies, and WHO staff responding to the pandemic H1N1 2009. Because of escalating outbreaks in several countries, other invited participants were unable to attend. This report summarizes discussions held at the meeting and presents some preliminary results of ongoing modelling of the pandemic H1N1 2009.
Public health priorities of the modelling network
Past pandemic plans have made use of results from mathematical modelling. However, because these plans were based on scenarios of more severe pandemics than the current one, revision is required. In consultation with WHO and public health ministries in several affected countries, the network has decided that its public health priorities are to:
Transmission dynamics of the H1N1 pandemic 2009
Several presentations at the meeting characterized the transmission dynamics of the pandemic virus, providing preliminary estimates of the basic reproduction number (R0),1 secondary attack rates, incubation period2 and generation time.3 Presentations also focused on epidemics in Australia; China, Hong Kong, Special Administrative Region (SAR); Japan; and several countries in Europe and North America. Although most presentations described national-scale epidemics, some work also focused on smaller-scale transmission within subnational regions (for example, Victoria, Australia) or localized outbreaks (for example, within schools).
Most analyses of data from Europe and the USA suggest that R0 is most likely to be 1.2?1.7, with higher estimates in specific contexts (for example, in Japan R0=2.3; 95% confidence interval [CI], 2.0?2.6; in New Zealand R0=1.96; 95% CI, 1.80?2.15; in Victoria, Australia, R0 was initially >2 but dropped during the containment phase; in Argentina and Chile R0>1.7). Differences in R0, even in this range, may have a profound impact on the extent to which interventions can control an epidemic (Fig. 1).
Groups from the United Kingdom and the USA presented estimates of generation time that were fairly consistent across groups and settings, with agreement that the mean generation time of the pandemic H1N1 2009 is 2.5?3 days. Fewer data were available regarding the median incubation period, but 1 estimate suggested that the incubation period is 1.4 days (95% CI, 1.0?1.8), which is similar to previously circulating influenza strains. These estimates have implications for estimating transmission parameters (for example, estimates of R0 derived from epidemic curves depend on the generation time assumed) and for public health (for example, the incubation period informs quarantine periods).
Analyses of household secondary attack rates from Hong Kong SAR, Italy, Mexico, the United Kingdom and the USA were reasonably consistent, at 18?30%. Attack rates in schools in Japan were low, ranging from <1?5.3%, (H. Nishiura, personal communication).
The influence of seasonality on transmission dynamics was identified as being poorly understood but an important issue for the network. As the pandemic progresses in countries in the southern hemisphere, it is important to collect accurate data on the growth of the epidemic and household attack rates to identify differences in the transmission of the pandemic H1N1 2009.
Characterizing severity
The severity of a pandemic virus can be evaluated from 2 perspectives: that of the individual who has been infected and from the population level ? that is, how many complications and deaths might be expected as a whole. The individual perspective is related to the pathogenicity of the virus and host determinants in affected patients, such as the proportion who are at increased risk of developing severe disease (for example, people with underlying conditions such as asthma, obesity, pregnancy), and it is related to whether there is access to appropriate medical treatment.
Population severity is a function of individual pathogenicity and the clinical attack rate,4 which may be affected by measures limiting transmission, such as the widespread use of antivirals or implementation of effective non-pharmaceutical interventions. Measuring severity from either perspective in real time is a major challenge, yet is of interest to policy-makers.
The most common measures of severity from the individual?s perspective are the case-fatality rate (CFR) and the hospitalization rate. However, there are challenges associated with estimating these rates in real time. One problem is that the denominator (the number of cases) may be underestimated since subclinical or asymptomatic cases will be missed if surveillance focuses on severe cases, leading to inflated estimates of hospitalization (if patients are hospitalized for illness rather than isolation) and CFRs. Furthermore, there may be delays before accurate discharge diagnoses are available.
A population perspective on pandemic severity requires estimates of the clinical attack rate. Models can give an upper bound on attack rates if an accurate estimate of R0 is available, but such models provide estimates of the serological attack rate (the proportion infected), rather than the proportion of those infected who display clinical symptoms.
It is not possible to measure the symptomatic fraction of all infections in real time. While serological testing of cohorts may provide this information, sensitive and specific serological tests for the human infection with pandemic A (H1N1) influenza virus are still under development, and useful results even from validated serological testing would be available only after the peak of the epidemic.
Alternatively, investigation of community outbreaks, focusing on the first populations affected, may provide estimates; these cases should ideally be confirmed by a laboratory.
Summary measures of severity may hide potentially important heterogeneity (by age group or risk group, for instance). Although determining the role of such co-factors is more challenging, it should be done to enable evidencebased prioritization of resources. The outbreak in La Gloria, Mexico, gives an example of a community where the highest clinical attack rate appeared to be among children aged ≤15 years, although only 1 case was confirmed as human infection with pandemic A (H1N1) influenza virus, and other influenza viruses were isolated from this community concurrently.
The impact of pharmaceutical and non-pharmaceutical interventions
Non-pharmaceutical interventions
For the vast majority of the world?s population, pharmaceutical interventions, such as vaccines and antivirals, will not have a significant role in the current pandemic. For most individuals living in low-resource settings, such pharmaceuticals likely will be in short supply, not available during the peak of the pandemic or too costly for widespread use. Therefore, any coordinated public- health effort to reduce the impact of the pandemic must rely largely on the combined effect of non-pharmaceutical interventions, such as school closures, restrictions on mass gatherings, withdrawal of symptomatic individuals to the home, and increased use of personal protective measures (handwashing and, perhaps, mask wearing). However, it needs to be emphasized that if populations do not have eventual access to pharmaceutical interventions, then public-health measures can achieve only modest reductions in attack rates overall: the epidemic will stop only when enough of the population has been infected to achieve herd immunity (so R0<1). Thus, non-pharmaceutical interventions may be too effective: if they reduce transmission so that herd immunity is not reached, then when they are lifted transmission will resume.
Initial observations suggest that the proportion of transmission occurring in school-age children is higher than expected, based on studies of seasonal influenza. A comparison of the age distribution of early pandemic H1N1 2009 cases in Hong Kong SAR and the USA suggests that closing kindergartens and primary schools may have had a considerable effect in reducing transmission (S. Riley, personal communication). In many populations, the economic and community costs associated with school closures are substantial. The timing of school closures is also a consideration. Modelling work suggests that for school closures to have the maximum effect on overall attack rates, schools must be closed before 1% of the population becomes ill. But substantial reductions in peak attack rates (and thus reductions in demands on health-care systems during peak periods) might be achieved with relatively short periods of school closure.
Therefore, in this pandemic, which has been mild so far, models may be of immediate use in identifying efficient policies for school closure that achieve the optimal balance of benefit in reducing
and slowing transmission with the minimum social disruption.
Improving hand hygiene and increasing mask wearing, if proven effective and if compliance is high, should be emphasized alongside the closure of schools and isolation of symptomatic cases, so that infections are genuinely averted and not just transferred from schools and workplaces to the home. However, empirical studies of the efficacy of these interventions, which recruited infectious individuals at primary health-care clinics, have produced equivocal results. The limited efficacy may be due to late application of the interventions in the trials, and these results do not
exclude high efficacy for increased hand hygiene and mask wearing if implemented earlier. In the short term, mathematical models could help estimate the per-unit-time efficacy of these interventions using existing data.
Vaccination
A number of modelling studies on vaccine strategy for pandemic influenza have been published. The moderate severity of the current pandemic offers additional vaccination challenges, some of which may be informed by modelling. If larger stockpiles of vaccines are available before the pandemic starts to spread, the optimal way of reducing morbidity and mortality is to reduce transmission and achieve herd immunity. Children are generally more likely to spread influenza than adults. Hence, targeting children for early vaccination may reduce morbidity and mortality through a combination of direct protection and reduced transmission.
Different countries will be in different positions regarding the amount of pandemic H1N1 2009 vaccine in their possession and the timing of delivery of the vaccines, their population structure, and cultural habits affecting social mixing (Fig. 2). When vaccine supplies are limited and immunization occurs in the face of a pandemic, an alternative strategy to targeting children is to target high-risk groups who are likely to develop severe disease. The trade-off between reducing transmission and targeting high-risk groups depends on whether herd immunity can realistically be achieved. Preliminary results presented at the meeting suggest that coverage of 50?70% of children will be necessary to achieve herd immunity and significantly reduce transmission, coverage levels that participants agreed would be hard to achieve for most countries. However, the high-risk groups in the current pandemic include young children, who are the most efficient at transmitting influenza viruses. Thus, even if achieving vaccination rates high enough to result in indirect protection through population, immunity may be unrealistic, targeting vaccine at children addresses concerns both of decreasing transmission and protecting directly those at greatest risk of serious complications from pandemic H1N1 2009 infection.
There are several uncertainties around pandemic H1N1 2009 vaccines, including the immunogenicity of the vaccines entering commercial production, the number of doses likely to be needed for full immunity in children and young adults, and the time required between doses, which may be crucial within the time frame of roll-out and the progression of the epidemic (Fig. 2). Additionally, vaccines may have different effects on immunized individuals; they may reduce susceptibility, infectiousness or severity. Modelling may provide important insights to aid decisionmaking in response to these issues.
Antiviral resistance
Many countries have stockpiled antivirals (oseltamivir or zanamivir) as part of their containment and mitigation strategies. Resistance against these antivirals could emerge due to point mutations of the pandemic strain or genetic reassortment among the pandemic strain and circulating strains of resistant seasonal influenza. As of 27 July 2009, oseltamivir resistance has been detected in 6 cases of human infection with pandemic A (H1N1) influenza virus: 5 cases (in Canada, Denmark and Japan) treated with oseltamivir and 1 (Hong Kong SAR) not treated with oseltamivir, indicating possible transmission of oseltamivir-resistant strains. This raises the concern that resistance could become widespread. If transmissibility of the detected resistant strain is comparable to that of wildtype virus, the prophylactic and therapeutic value of oseltamivir stockpiles will be substantially reduced. Surveillance for antiviral resistance among currently circulating influenza viruses is a high priority for the WHO influenza coordinating centres.
Modelling has previously been used to assess the potential impact of the emergence and spread of resistance during large-scale antiviral interventions used for an influenza pandemic and to devise strategies to limit resistance. Similar modelling techniques could be used for real-time surveillance of resistance during the pandemic H1N1 2009.
Discussion
During the next few months the pandemic H1N1 2009 will pose challenges across the world: how to make best use of limited vaccine and antiviral supplies; how to manage demands on health-care resources, which may peak substantially above normal capacity; and how to determine the role that non-pharmaceutical interventions might have, given the observed level of severity and the economic disruption caused by such measures. Thus, while every country shares a common goal of minimizing mortality and morbidity, differences among countries ? in health-care capacity, access to vaccines and antivirals, and in the timing of the pandemic ? will require each country to tailor its own response. A particular priority for WHO?s modelling network is to adapt the experiences and lessons learnt from pandemic planning in developed countries to the needs of developing and transitional countries.
During the meeting, work highlighted the insights into transmission and control that rapid-response modelling can provide. However, participants agreed that fundamental uncertainties (such as the impact of seasonal variation in transmissibility of the influenza virus) made prediction of the likely trajectory of the pandemic uncertain. The limitations of surveillance data make understanding the current situation and short-term extrapolation of recent trends the limit of what modelling may provide in terms of prediction, at least until the pandemic has peaked in a country.
The network agreed that well designed cross-sectional serological surveys are needed to document the actual extent of infection in various populations. In addition, cohort studies, sentinel surveillance and hospital-based syndromic surveillance are vital to track the epidemic in different countries and provide policy-makers with the situational awareness required to make evidence-based decisions.
In considering the intervention options open to policymakers, the network recognized the many constraints under which governments and public-health agencies are acting (Fig. 3). For example, levels of antiviral supplies are largely fixed, and vaccine production capacity has been commissioned, so the only flexibility is in determining the groups who should be prioritized to receive vaccination.
The majority of the world?s countries will have limited supplies early on: health-care workers and others deemed critical to the pandemic response system will be the highest priority for vaccination. Only a few developed countries will be able to use vaccine to minimize transmission; in those cases, modelling will determine the risks and benefits of such a strategy. Based on the epidemiological data, children will be a high priority for vaccination both from the perspective of possibly reducing transmission and from the individual-level perspective, given the higher attack rates in children and younger adults when compared with older adults.
More flexibility exists in the range of non-pharmaceutical interventions. Balancing the limited or unknown benefit of different measures against their disruptive effect on society is crucial, especially in the context of a moderate pandemic. This issue is particularly important in developing countries that have limited access to vaccines and where non-pharmaceutical interventions will have limited impact on the overall clinical attack rate or mortality. The modelling of these issues in the next weeks and months will be invaluable in identifying public-health measures appropriate for developing countries ? for example, using the targeted closure of schools around the peak of the pandemic to reduce peak incidence and thus demand on limited health-care resources.
All members of the network agreed that the network could yield longer-term value in building global capacity in modelling and epidemiological analysis as well as a base of educated consumers of modelling in public-health agencies across the world. In particular, the current crisis offers an unrivalled opportunity to establish long-term collaborative links among academic centres of expertise in modelling and public-health agencies in developing and transitional countries. Participants at the meeting agreed that in order for modelling to influence decisions and policy, it is essential that modellers build long-term trust and collaborative relationships with public-health agencies.
-
1 R0: the average number of secondary cases generated from 1 case at the start of an epidemic.
2 Incubation period: the time between infection and symptom onset.
3 Generation time: the mean delay between the time of infection of an index case and the times of infection of secondary cases infected by the index case.
4 Clinical attack rate: the proportion of the population infected and showing clinical symptoms.
-
-----
[Original Full Document: LINK. EDITED.]
Weekly epidemiological record - 21 august 2009, 84rd year - No. 34, 2009, 84, 341?352 - http://www.who.int/wer
Mathematical modelling of the pandemic H1N1 2009
Background and objectives of an informal network coordinated by WHO
Mathematical and computational models are useful tools for elucidating the non-linear transmission dynamics of epidemics, for analysing epidemiological data and for examining the potential impact of possible intervention options. Real-time mathematical modelling from an informal network of academic mathematical modellers coordinated by WHO during the severe acute respiratory syndrome (SARS) epidemic in 2003 yielded valuable insights, which informed strategy and control measures.
Soon after WHO declared that the emergence of a novel influenza A (H1N1) strain in Mexico and the United States represented a ?public health emergency of international concern?, WHO convened another informal network of mathematical modellers with the following goals:
(a) to describe and predict the behaviour and impact of the pandemic H1N1 2009 and demonstrate the potential outcome of proposed pharmaceutical and non-pharmaceutical interventions in different settings;
(b) to present these analyses in formats suitable for various audiences, including technical experts, policy-makers and the general public; and
(c) to adapt models and interpret experiences from developed countries so they can be applied in low-resource countries.
The mathematical modelling network met in Geneva in early July 2009. The network consists of representatives from infectious disease modelling groups (academic and public health institutions), professionals from national public health agencies, and WHO staff responding to the pandemic H1N1 2009. Because of escalating outbreaks in several countries, other invited participants were unable to attend. This report summarizes discussions held at the meeting and presents some preliminary results of ongoing modelling of the pandemic H1N1 2009.
Public health priorities of the modelling network
Past pandemic plans have made use of results from mathematical modelling. However, because these plans were based on scenarios of more severe pandemics than the current one, revision is required. In consultation with WHO and public health ministries in several affected countries, the network has decided that its public health priorities are to:
- better understand the transmission dynamics of the pandemic virus in various settings (for example, estimate the reproductive number [R0] and secondary attack rates in schools, households and more widely in communities in affected countries), and better understand the effects of seasonality on the growth of the epidemic in the Northern and Southern Hemispheres;
- characterize severity at an individual level and a population level; and
- evaluate the impacts of pharmaceutical interventions (vaccines, antivirals) and non-pharmaceutical interventions (for example, school closures, mask wearing) on transmission and severity in different settings.
Transmission dynamics of the H1N1 pandemic 2009
Several presentations at the meeting characterized the transmission dynamics of the pandemic virus, providing preliminary estimates of the basic reproduction number (R0),1 secondary attack rates, incubation period2 and generation time.3 Presentations also focused on epidemics in Australia; China, Hong Kong, Special Administrative Region (SAR); Japan; and several countries in Europe and North America. Although most presentations described national-scale epidemics, some work also focused on smaller-scale transmission within subnational regions (for example, Victoria, Australia) or localized outbreaks (for example, within schools).
Most analyses of data from Europe and the USA suggest that R0 is most likely to be 1.2?1.7, with higher estimates in specific contexts (for example, in Japan R0=2.3; 95% confidence interval [CI], 2.0?2.6; in New Zealand R0=1.96; 95% CI, 1.80?2.15; in Victoria, Australia, R0 was initially >2 but dropped during the containment phase; in Argentina and Chile R0>1.7). Differences in R0, even in this range, may have a profound impact on the extent to which interventions can control an epidemic (Fig. 1).
Groups from the United Kingdom and the USA presented estimates of generation time that were fairly consistent across groups and settings, with agreement that the mean generation time of the pandemic H1N1 2009 is 2.5?3 days. Fewer data were available regarding the median incubation period, but 1 estimate suggested that the incubation period is 1.4 days (95% CI, 1.0?1.8), which is similar to previously circulating influenza strains. These estimates have implications for estimating transmission parameters (for example, estimates of R0 derived from epidemic curves depend on the generation time assumed) and for public health (for example, the incubation period informs quarantine periods).
Analyses of household secondary attack rates from Hong Kong SAR, Italy, Mexico, the United Kingdom and the USA were reasonably consistent, at 18?30%. Attack rates in schools in Japan were low, ranging from <1?5.3%, (H. Nishiura, personal communication).
The influence of seasonality on transmission dynamics was identified as being poorly understood but an important issue for the network. As the pandemic progresses in countries in the southern hemisphere, it is important to collect accurate data on the growth of the epidemic and household attack rates to identify differences in the transmission of the pandemic H1N1 2009.
Characterizing severity
The severity of a pandemic virus can be evaluated from 2 perspectives: that of the individual who has been infected and from the population level ? that is, how many complications and deaths might be expected as a whole. The individual perspective is related to the pathogenicity of the virus and host determinants in affected patients, such as the proportion who are at increased risk of developing severe disease (for example, people with underlying conditions such as asthma, obesity, pregnancy), and it is related to whether there is access to appropriate medical treatment.
Population severity is a function of individual pathogenicity and the clinical attack rate,4 which may be affected by measures limiting transmission, such as the widespread use of antivirals or implementation of effective non-pharmaceutical interventions. Measuring severity from either perspective in real time is a major challenge, yet is of interest to policy-makers.
The most common measures of severity from the individual?s perspective are the case-fatality rate (CFR) and the hospitalization rate. However, there are challenges associated with estimating these rates in real time. One problem is that the denominator (the number of cases) may be underestimated since subclinical or asymptomatic cases will be missed if surveillance focuses on severe cases, leading to inflated estimates of hospitalization (if patients are hospitalized for illness rather than isolation) and CFRs. Furthermore, there may be delays before accurate discharge diagnoses are available.
A population perspective on pandemic severity requires estimates of the clinical attack rate. Models can give an upper bound on attack rates if an accurate estimate of R0 is available, but such models provide estimates of the serological attack rate (the proportion infected), rather than the proportion of those infected who display clinical symptoms.
It is not possible to measure the symptomatic fraction of all infections in real time. While serological testing of cohorts may provide this information, sensitive and specific serological tests for the human infection with pandemic A (H1N1) influenza virus are still under development, and useful results even from validated serological testing would be available only after the peak of the epidemic.
Alternatively, investigation of community outbreaks, focusing on the first populations affected, may provide estimates; these cases should ideally be confirmed by a laboratory.
Summary measures of severity may hide potentially important heterogeneity (by age group or risk group, for instance). Although determining the role of such co-factors is more challenging, it should be done to enable evidencebased prioritization of resources. The outbreak in La Gloria, Mexico, gives an example of a community where the highest clinical attack rate appeared to be among children aged ≤15 years, although only 1 case was confirmed as human infection with pandemic A (H1N1) influenza virus, and other influenza viruses were isolated from this community concurrently.
The impact of pharmaceutical and non-pharmaceutical interventions
Non-pharmaceutical interventions
For the vast majority of the world?s population, pharmaceutical interventions, such as vaccines and antivirals, will not have a significant role in the current pandemic. For most individuals living in low-resource settings, such pharmaceuticals likely will be in short supply, not available during the peak of the pandemic or too costly for widespread use. Therefore, any coordinated public- health effort to reduce the impact of the pandemic must rely largely on the combined effect of non-pharmaceutical interventions, such as school closures, restrictions on mass gatherings, withdrawal of symptomatic individuals to the home, and increased use of personal protective measures (handwashing and, perhaps, mask wearing). However, it needs to be emphasized that if populations do not have eventual access to pharmaceutical interventions, then public-health measures can achieve only modest reductions in attack rates overall: the epidemic will stop only when enough of the population has been infected to achieve herd immunity (so R0<1). Thus, non-pharmaceutical interventions may be too effective: if they reduce transmission so that herd immunity is not reached, then when they are lifted transmission will resume.
Initial observations suggest that the proportion of transmission occurring in school-age children is higher than expected, based on studies of seasonal influenza. A comparison of the age distribution of early pandemic H1N1 2009 cases in Hong Kong SAR and the USA suggests that closing kindergartens and primary schools may have had a considerable effect in reducing transmission (S. Riley, personal communication). In many populations, the economic and community costs associated with school closures are substantial. The timing of school closures is also a consideration. Modelling work suggests that for school closures to have the maximum effect on overall attack rates, schools must be closed before 1% of the population becomes ill. But substantial reductions in peak attack rates (and thus reductions in demands on health-care systems during peak periods) might be achieved with relatively short periods of school closure.
Therefore, in this pandemic, which has been mild so far, models may be of immediate use in identifying efficient policies for school closure that achieve the optimal balance of benefit in reducing
and slowing transmission with the minimum social disruption.
Improving hand hygiene and increasing mask wearing, if proven effective and if compliance is high, should be emphasized alongside the closure of schools and isolation of symptomatic cases, so that infections are genuinely averted and not just transferred from schools and workplaces to the home. However, empirical studies of the efficacy of these interventions, which recruited infectious individuals at primary health-care clinics, have produced equivocal results. The limited efficacy may be due to late application of the interventions in the trials, and these results do not
exclude high efficacy for increased hand hygiene and mask wearing if implemented earlier. In the short term, mathematical models could help estimate the per-unit-time efficacy of these interventions using existing data.
Vaccination
A number of modelling studies on vaccine strategy for pandemic influenza have been published. The moderate severity of the current pandemic offers additional vaccination challenges, some of which may be informed by modelling. If larger stockpiles of vaccines are available before the pandemic starts to spread, the optimal way of reducing morbidity and mortality is to reduce transmission and achieve herd immunity. Children are generally more likely to spread influenza than adults. Hence, targeting children for early vaccination may reduce morbidity and mortality through a combination of direct protection and reduced transmission.
Different countries will be in different positions regarding the amount of pandemic H1N1 2009 vaccine in their possession and the timing of delivery of the vaccines, their population structure, and cultural habits affecting social mixing (Fig. 2). When vaccine supplies are limited and immunization occurs in the face of a pandemic, an alternative strategy to targeting children is to target high-risk groups who are likely to develop severe disease. The trade-off between reducing transmission and targeting high-risk groups depends on whether herd immunity can realistically be achieved. Preliminary results presented at the meeting suggest that coverage of 50?70% of children will be necessary to achieve herd immunity and significantly reduce transmission, coverage levels that participants agreed would be hard to achieve for most countries. However, the high-risk groups in the current pandemic include young children, who are the most efficient at transmitting influenza viruses. Thus, even if achieving vaccination rates high enough to result in indirect protection through population, immunity may be unrealistic, targeting vaccine at children addresses concerns both of decreasing transmission and protecting directly those at greatest risk of serious complications from pandemic H1N1 2009 infection.
There are several uncertainties around pandemic H1N1 2009 vaccines, including the immunogenicity of the vaccines entering commercial production, the number of doses likely to be needed for full immunity in children and young adults, and the time required between doses, which may be crucial within the time frame of roll-out and the progression of the epidemic (Fig. 2). Additionally, vaccines may have different effects on immunized individuals; they may reduce susceptibility, infectiousness or severity. Modelling may provide important insights to aid decisionmaking in response to these issues.
Antiviral resistance
Many countries have stockpiled antivirals (oseltamivir or zanamivir) as part of their containment and mitigation strategies. Resistance against these antivirals could emerge due to point mutations of the pandemic strain or genetic reassortment among the pandemic strain and circulating strains of resistant seasonal influenza. As of 27 July 2009, oseltamivir resistance has been detected in 6 cases of human infection with pandemic A (H1N1) influenza virus: 5 cases (in Canada, Denmark and Japan) treated with oseltamivir and 1 (Hong Kong SAR) not treated with oseltamivir, indicating possible transmission of oseltamivir-resistant strains. This raises the concern that resistance could become widespread. If transmissibility of the detected resistant strain is comparable to that of wildtype virus, the prophylactic and therapeutic value of oseltamivir stockpiles will be substantially reduced. Surveillance for antiviral resistance among currently circulating influenza viruses is a high priority for the WHO influenza coordinating centres.
Modelling has previously been used to assess the potential impact of the emergence and spread of resistance during large-scale antiviral interventions used for an influenza pandemic and to devise strategies to limit resistance. Similar modelling techniques could be used for real-time surveillance of resistance during the pandemic H1N1 2009.
Discussion
During the next few months the pandemic H1N1 2009 will pose challenges across the world: how to make best use of limited vaccine and antiviral supplies; how to manage demands on health-care resources, which may peak substantially above normal capacity; and how to determine the role that non-pharmaceutical interventions might have, given the observed level of severity and the economic disruption caused by such measures. Thus, while every country shares a common goal of minimizing mortality and morbidity, differences among countries ? in health-care capacity, access to vaccines and antivirals, and in the timing of the pandemic ? will require each country to tailor its own response. A particular priority for WHO?s modelling network is to adapt the experiences and lessons learnt from pandemic planning in developed countries to the needs of developing and transitional countries.
During the meeting, work highlighted the insights into transmission and control that rapid-response modelling can provide. However, participants agreed that fundamental uncertainties (such as the impact of seasonal variation in transmissibility of the influenza virus) made prediction of the likely trajectory of the pandemic uncertain. The limitations of surveillance data make understanding the current situation and short-term extrapolation of recent trends the limit of what modelling may provide in terms of prediction, at least until the pandemic has peaked in a country.
The network agreed that well designed cross-sectional serological surveys are needed to document the actual extent of infection in various populations. In addition, cohort studies, sentinel surveillance and hospital-based syndromic surveillance are vital to track the epidemic in different countries and provide policy-makers with the situational awareness required to make evidence-based decisions.
In considering the intervention options open to policymakers, the network recognized the many constraints under which governments and public-health agencies are acting (Fig. 3). For example, levels of antiviral supplies are largely fixed, and vaccine production capacity has been commissioned, so the only flexibility is in determining the groups who should be prioritized to receive vaccination.
The majority of the world?s countries will have limited supplies early on: health-care workers and others deemed critical to the pandemic response system will be the highest priority for vaccination. Only a few developed countries will be able to use vaccine to minimize transmission; in those cases, modelling will determine the risks and benefits of such a strategy. Based on the epidemiological data, children will be a high priority for vaccination both from the perspective of possibly reducing transmission and from the individual-level perspective, given the higher attack rates in children and younger adults when compared with older adults.
More flexibility exists in the range of non-pharmaceutical interventions. Balancing the limited or unknown benefit of different measures against their disruptive effect on society is crucial, especially in the context of a moderate pandemic. This issue is particularly important in developing countries that have limited access to vaccines and where non-pharmaceutical interventions will have limited impact on the overall clinical attack rate or mortality. The modelling of these issues in the next weeks and months will be invaluable in identifying public-health measures appropriate for developing countries ? for example, using the targeted closure of schools around the peak of the pandemic to reduce peak incidence and thus demand on limited health-care resources.
All members of the network agreed that the network could yield longer-term value in building global capacity in modelling and epidemiological analysis as well as a base of educated consumers of modelling in public-health agencies across the world. In particular, the current crisis offers an unrivalled opportunity to establish long-term collaborative links among academic centres of expertise in modelling and public-health agencies in developing and transitional countries. Participants at the meeting agreed that in order for modelling to influence decisions and policy, it is essential that modellers build long-term trust and collaborative relationships with public-health agencies.
-
1 R0: the average number of secondary cases generated from 1 case at the start of an epidemic.
2 Incubation period: the time between infection and symptom onset.
3 Generation time: the mean delay between the time of infection of an index case and the times of infection of secondary cases infected by the index case.
4 Clinical attack rate: the proportion of the population infected and showing clinical symptoms.
-
-----
Comment