Nancy J. Cox

The Centers for Disease Control and Prevention (CDC) and other groups participating in the World Health Organization's (WHO) global influenza surveillance network track the ever-changing influenza viruses that infect humans. But influenza viruses have characteristics that make this task difficult. One of these characteristics is their ability to evolve relatively rapidly. Another factor is their ability to infect avian and other mammalian species and occasionally cross host species' barriers and evolve rapidly in a new host. Expanding global influenza surveillance, in conjunction with automating key laboratories and establishing an informatics infrastructure, could dramatically improve our ability to track influenza viruses and prevent and control influenza outbreaks in the future.

The most widely recognized consequence of influenza is the morbidity and mortality due to pandemics (worldwide epidemics). For example, more than 500,000 excess influenza-related deaths occurred in 1918 in the United States during the “Spanish” influenza pandemic alone, and a total of more than 600,000 deaths have been attributed to the Spanish, Asian (1957), and Hong Kong (1968) influenza pandemics combined. Furthermore, the morbidity due to pandemics is well recognized to cause considerable social disruption and substantial economic losses. It is less well

recognized that the cumulative totals for mortality and morbidity are actually much greater during interpandemic periods than pandemic years. For just the quarter century between 1970 and 1995, there were also more than 600,000 excess influenza-related deaths. During the current interpandemic period, an annual average of 20,000 excess influenza-related deaths and more than 100,000 excess influenza-related hospitalizations have occurred since 1973. In addition, serious epidemics have resulted in more than 40,000 influenza-related deaths and 200,000 influenza-related hospitalizations.

Influenza affects different age groups in different ways. During the current interpandemic period, more than 90 percent of the excess mortality from pneumonia and influenza occurs in the elderly. Thus, the aging of the population, one of the demographic changes occurring worldwide, has significant public health implications with regard to influenza. Between 1985 and 2025, the population age 65 years and older in the United States will double. In China and India this population group will triple in size. If effective influenza prevention and control measures are not instituted, the world will suffer from substantial societal and economic costs related to influenza, particularly among older individuals. Although mortality during the current interpandemic period occurs primarily among those 65 and older, influenza-related hospitalizations occur in both the very young and the elderly. School-aged and younger children have the highest rates of medically attended influenza-related illness.

During the three pandemics of influenza that occurred during the twentieth century, the pattern of mortality differed from that seen today. During the Spanish influenza pandemic of 1918, more than 99 percent of excess pneumonia and influenza deaths occurred among persons under age 65. However, as the new pandemic viruses became established in the human population and continued to circulate and evolve, a decrease was observed in the proportion of influenza-related deaths occurring in those under age 65. A similar pattern was observed for the pandemics of Asian influenza in 1957 and the Hong Kong influenza in 1968. When the Asian strain of influenza first emerged in humans, approximately 40 percent of influenza-related deaths occurred in those under age 65. During the 1968 pandemic, approximately 60 percent of the deaths occurred in this age group. However, smaller proportions of influenza-related deaths were observed among those under 65 years of age, as these two novel viruses continued to circulate among humans and population immunity developed.

Influenza exhibits a complex seasonal pattern of circulation. In temperate regions of the Northern Hemisphere, influenza activity usually peaks during the months of December, January, and February. In temperate regions of the Southern Hemisphere, the peak of influenza activity occurs from June through August. In tropical regions, influenza can be isolated almost year round, and in some areas there are two peaks of influenza activity. This pattern of influenza circulation keeps WHO's global influenza surveillance network busy throughout the year. This network, established by the WHO over 50 years ago, has grown to include four WHO collaborating centers for influenza (CCIs) located in London, Atlanta, Melbourne, and Tokyo. It also includes approximately 110 national influenza centers (NICs) in over 80 countries worldwide. These centers are fairly well distributed around the world; however, some countries in Africa, South America, and the Middle East currently do not participate in WHO's global influenza surveillance network.

Active NICs isolate influenza viruses and identify them using a WHO influenza reagent kit that is prepared annually at the CDC and distributed throughout the world. Subsequently, the isolated viruses are sent to one or more CCIs to be analyzed in greater detail. The purpose of this virological surveillance system is to detect a variant strain that could cause the next epidemic or pandemic of influenza with sufficient early warning to include the new strain in influenza vaccine. Epidemiologic information is also collected and sent to WHO in an effort to document the extent of influenza activity during the time the viruses were isolated. In addition, the United States (and a number of other countries) has its own domestic network, which includes more than 70 WHO collaborating laboratories, many of which are located in state health departments.

The four CCIs have several key responsibilities, including the following: analyzing influenza viruses from around the world; communicating with the NICs; providing training; communicating with other WHO collaborating centers; preparing a WHO influenza reagent kit; and providing data for WHO's annual vaccine recommendations. All four collaborating centers use antigenic and genetic techniques to comparatively analyze the influenza viruses received. The CDC has also provided extensive support for laboratory training. For example, in 1994 and 1999 training in Beijing was provided for participants from municipal and provincial antiepidemic stations across China. The CCIs communicate with each other, with the NICs, and with WHO headquarters in Geneva. Much of this information is published in WHO's Weekly Epidemiological Record.

The four WHO collaborating centers for influenza are responsible for providing data for the influenza vaccine recommendations that WHO issues to the world twice a year—in February for the Northern Hemisphere and in September for the Southern Hemisphere. These recommendations are based on antigenic, genetic, serological, and epidemiologic data.

After viruses are received by the CCIs, they are amplified by growth in tissue culture or embryonated hens' eggs. A hemagglutination inhibition (HI) test—a serological assay that can rapidly screen a large number of viruses—is used to detect new antigenic variants. This test is performed using postinfection ferret sera made against reference strains. These sera are very sensitive for detecting antigenic differences among viruses. Low reactors, that is, those viruses that are not well inhibited by the reference ferret antisera and appear to be variant viruses, are retested, and viruses that are confirmed to be variants are inoculated into ferrets to produce a corresponding antiserum for further HI testing.

The WHO relies on ministries of health and the NCIs to report high levels of influenza-like activity coincident with isolation of the variant strains. Before selecting strains for inclusion in the vaccine, the WHO also determines if a reduced postvaccine immune response to variant viruses occurs in individuals who have received the current vaccine. After all of this information is synthesized, recommendations are made regarding the specific viruses that should be included in the next year's influenza vaccine. Once the vaccine strain recommendations are issued, suitable candidate vaccine strains are distributed to the vaccine manufacturers by the WHO collaborating centers or by regulatory authorities.

In 1985 the CDC began prospective sequencing of hemagglutinin (HA) genes to determine if the data generated would be useful for vaccine strain selection. In recent years the CDC has sequenced the HA genes of many influenza field strains, particularly those that may have an altered pattern of reactivity in the HI test. CDC has focused on the HA gene because antibodies to hemagglutinin determine whether or not a person is protected against infection by influenza. When CDC began prospective sequencing of the HA genes of influenza field strains it was not known if sequencing would be a useful adjunct to reference serological analysis for vaccine strain selection; however, it is now well established that molecular methods such as sequencing are extremely useful. Sequence analysis of HA genes can provide information concerning the molecular basis for antigenic drift, an area where more information is needed. We calculate rates of change for the HA at both the nucleotide and amino acid levels and also examine the types of amino acid changes in the hemagglutinin that appear to confer antigenic changes and where those changes are

located in the three-dimensional structure of the molecule. We are also interested in learning if there are predictable patterns of change. Information gained from ongoing sequencing indicates that the number of pathways for evolution of the virus may be limited and that there may be some predictability in the patterns of change. It may be possible to identify specific clues in the sequence data and devise methods to help predict which strains are most likely to cause epidemics in the future. The CDC has also been working with scientists at the Los Alamos National Laboratory to establish an international database for influenza gene sequence data.

Table 4.1 shows the “WHO report card,” indicating the degree of antigenic relatedness or “match” between the epidemic strains that have circulated in the United States over the past 10 years and the corresponding

vaccine strains. This table shows that despite a good record overall there was a poor match between the vaccine strain and the circulating strains during the 1997 to 1998 influenza season. Because the A/Sydney/5/97 variant was not detected until the summer of 1997, there was not enough time to include it in the vaccine before the following influenza season. Despite this mismatch, the similarity between the antigenic properties of the vaccine strains and those of the predominant circulating viruses has improved during the past 10 years compared to previous decades. Factors that have contributed significantly to an improved match include better worldwide surveillance, especially in some European countries; the expansion of influenza surveillance in China; and the use of molecular analysis as an adjunct to serological analysis.

Over the past 10 or 15 years, influenza vaccination rates have improved dramatically in certain segments of the U.S. population. Vaccination rates among people 65 and older have increased from approximately 33 percent in 1989 to 63 percent in 1997. Thus, the Healthy People 2000 Public Health Service objective of vaccinating 60 percent of this population was reached before the year 2000. However, vaccination rates have remained low among people under age 65 with high-risk conditions; more effort must be focused on improving vaccine coverage in this group.

WHO collaborating centers for Influenza face a number of limiting factors that affect vaccine strain selection as well as special investigations like those that occurred in Hong Kong after human infections with avian influenza A (H5N1) and (H9N2) viruses were documented. During the H5N1 investigation, it was necessary to test more than 3,000 sera using multiple tests and to repeat many assays. Development and modification of confirmatory assays were essential components of this investigation. Also, the entire genomes of the H5N1 influenza viruses isolated from humans needed to be sequenced quickly in order to try to determine why these viruses were able to jump the host-species barrier. Automation of these processes would have aided this work.

There are also limiting factors for generating data for influenza vaccine strain selection. Influenza surveillance is not uniform throughout the world, and in some countries there is a need to establish influenza surveillance or to increase the number of influenza viruses that are isolated. For example, the WHO network in certain parts of the world, including Africa, needs to be expanded. The CDC has been focusing on improving influenza surveillance in China, Vietnam, and Russia; influenza surveillance is also being enhanced in several countries in Asia, South America,

and Europe. In countries with very active NICs, many influenza isolates are often identified; however, not all of them are analyzed in detail. There is a need to increase the number of isolates sent to WHO CCIs for detailed analysis. This means that CCIs need to increase their capacity to grow, store, and analyze influenza strains and to devise improved methods to handle high throughput of influenza viruses at critical times. For example, in February, when the vaccine strain selection for the Northern Hemisphere is being made, and in September, when the vaccine strain selection for the Southern Hemisphere is being made, it would be extremely useful to have better ways to analyze large numbers of influenza isolates rapidly. There is also a need to expand the number of isolates sequenced and the number of genes for each virus. Furthermore, it would be extremely useful to extend serological testing of isolates by using a functional assay such as a neutralization test.

Routine surveillance efforts at several CCIs have reached maximum capacity, unless staffing is expanded significantly or processes are automated. Automation would greatly increase the number of influenza isolates and serum specimens that could be analyzed and would also increase the numbers and types of tests that could be performed. The new data generated would be extremely useful for extending our knowledge of the epidemiology, ecology, evolution, and pathogenesis of influenza viruses as well as for influenza vaccine strain selection. Efforts should be made to introduce automation into one or more of the WHO CCIs so that these additional data can be generated rapidly.