8.4 Control of microbial threats: population surveillance, vaccine studies, and the microbiology laboratory
Oxford Textbook of Public Health
Control of microbial threats: population surveillance, vaccine studies, and the microbiology laboratory
Norman D. Noah
Types of surveillance
Essentials of a surveillance system
Anticipating changes in incidence
Early detection of outbreaks
Evaluation of effectiveness of preventive measures
Identification of vulnerable groups
Setting priorities for allocation of resources
Field investigations of vaccines
Assessment of the need for a vaccine
Uptake and implementation of vaccines
The epidemiologist and the microbiology laboratory
Microbiology for epidemiologists
Epidemiology for microbiologists
Microbiological hazards were until only very recently the foremost causes of death and illness. Even though regulatory priority is now often given to chemical and radiation hazards, they remain serious and are again becoming common causes of ill health. Indeed, since the last edition of this textbook, much has been written on the theme of emerging or resurgent infectious disease. The epidemiological methods used in the investigation of infectious disease are not substantially different from those used in any epidemiological investigation. Nevertheless, there are some investigative techniques that are particularly applicable to infectious diseases. The techniques of surveillance were first used in infectious diseases and the special principles of surveillance as applied to infectious disease are described below in the first part of this chapter. Vaccine trials have many similarities to other epidemiological studies, but nevertheless are sufficiently distinctive to warrant special consideration: they are described in the second part of the chapter. The use of the microbiological laboratory in epidemiological studies of infectious disease, which is crucial to the discipline, is discussed in the third section. The investigation of the host, agent, and environment is described in Chapter 8.5.
The definition of population surveillance is:
Continuous analysis, interpretation, and feedback of systematically collected data, generally using methods distinguished by their practicability, uniformity and rapidity, rather than by accuracy or completeness. (Eylenbosch and Noah 1988; Last 1995).
Thus it is a type of observational study (Thacker et al. 1983). There are several key words in the definition as provided above which make this the preferred definition. ‘Continuous’ distinguishes surveillance from a survey. ‘Practicability’ is the essence of any workable surveillance system. ‘Uniformity’ ensures that the data can be interpreted sensibly, especially as surveillance is all about trends. ‘Rapidity’ is important for the system to be useful. And the words ‘rather than complete accuracy or completeness’ sum up the rough and ready philosophy behind most surveillance systems, which exist primarily for following trends in disease patterns—and for following them quickly. Surveillance essentially is not an academic process, although this does not mean that academic rigour is unnecessary in good practice. However, many of these points need to be qualified and are further discussed later in this section.
The word ‘monitoring’ should not be used interchangeably with surveillance: it is the ongoing evaluation of a control or management process (Eylenbosch and Noah 1988). The techniques of surveillance are usually necessary for efficient monitoring. Thus monitoring the success or failure of a vaccine policy for a disease will involve surveillance of vaccine use and surveillance of the disease, and possibly include surveillance of side-effects of vaccine and population immunity. Monitoring can become a finely tuned measure in which outcome can be constantly evaluated to adjust the process.
Research is not an essential part of surveillance, although surveillance may facilitate research. Consequently, there has been some discussion about the term ‘epidemiological surveillance’, and whether ‘public health surveillance’ is more valid (Thacker and Berkelman 1988).
Types of surveillance
Active and passive surveillance
In passive surveillance the recipient waits for the provider to report. In active surveillance routine checks of the provider are regularly made to ensure uniform and complete reporting. Enhanced surveillance is a term sometimes used to denote a form of active surveillance.
The globally successful smallpox eradication programme used active surveillance; each local health unit was ‘coerced, persuaded, and cajoled’ to report cases of smallpox each week, intensive further case finding was undertaken when a case was notified, and sources of information other than medical—teachers, schoolchildren, civil, and so on—were used (Henderson 1976). The reporting of a negative return is an important but not an essential part of a surveillance system. For surveillance of very rare diseases, however, completeness becomes more important, and ‘negative reporting’ becomes essential. Negative reporting has been used most effectively in the surveillance system for rare paediatric diseases, such as Reye’s syndrome and Kawasaki disease, run by the British Paediatric Surveillance Unit in Britain (Hall and Glickman 1988).
Sentinel surveillance is essentially a type of ‘sample surveillance’ in which reporting sources are situated at various sites covering an area (which may be very large, e.g. a country), and may provide fairly complete reporting within the population covered by each reporting source. This ensures that resources are not wasted in a large unwieldy surveillance system. If the population (total, or age and sex distributions) covered by each reporting centre is known, estimates of the disease or other indicator under surveillance can be made. A good discussion of some of the strengths and weaknesses of a sentinel system based on reporting by general practitioners (GPs) can be found in Boussard et al. (1996). Systems based on GP reporting are the most common forms of sentinel surveillance and are discussed later in this chapter.
Essentials of a surveillance system
The essential steps in any surveillance system are as follows:
the collection of data
the analysis of data
the interpretation of data
feedback of information.
These steps are similar to those taken in any scientific process. The collection of data is clearly the basic element of the system, but a failure of any of the other three steps in the process could also lead to failure of the system.
Collection of data
The collection of data has to be systematic, regular, and uniform, and, in infectious diseases particularly, topical and relevant. For collection to be systematic, suppliers of information should understand clearly what needs to be reported, leaving little scope for value judgements in deciding what to report, such as reporting only interesting or rare cases. With infectious disease surveillance, serious infections, rare infections, or those against which a control measure is available or is being planned, tend to be most worthwhile for surveillance.
For clinical reporting case definitions are helpful, especially for new or rare diseases, and even for common and easily recognizable diseases, particularly when they become rare. An early case definition for acquired immune deficiency syndrome (AIDS) was essential in the initial surveillance of this infection. Measles is easily recognizable clinically, but a case definition became essential—as indeed did laboratory confirmation of each case notified—when the vaccine campaign in the United States and in England became so successful that the infection became rare.
Case definitions may be rigid, but do not necessarily have to be highly specific. Careful consideration has to be given in individual instances as to whether sensitivity is more important than specificity. It may be important, for example, to choose a high-sensitive low-specific definition to encourage reporting, and then to concentrate on encouraging high specificity as the surveillance system becomes established.
For laboratory reporting, likewise, the criteria for what constitutes an acceptable report need to be understood clearly by the laboratory; for example, isolation of the organism from particular sites, or a fourfold or greater rise in antibody titre, or a single antibody titre above a certain level associated with a clinical feature characteristic of that infection, may all be acceptable, providing reporting is consistent. Nevertheless, in infectious disease there are often difficulties in associating a correctly identified organism with a particular illness, such as isolating an echovirus from a patient with gastroenteritis. In these instances all laboratories should understand clearly whether every such isolation need be reported, or only one where the isolate is considered to contribute to the symptoms (which may involve a value judgement). In the laboratory reporting system run by the Communicable Disease Surveillance Centre in England and Wales, all viral isolates are reported, although there is space on the form for the laboratory to indicate if the isolate is thought not to be relevant to the condition of the patient.
If feedback is to be regular, reporting must also be regular. The Communicable Disease Report for England and Wales is produced weekly, and the laboratories report weekly. Regular reporting also helps to maintain discipline and routine, essential ingredients of any surveillance system. The information received from providers must be relevant and topical, otherwise the interest of the participants is rapidly lost and their response will diminish.
Analysis of data
The analysis of data for infectious diseases is similar to that for any other type of epidemiological statistic. The basic principles of analysis by time, place, and person are fundamental.
Analysis by time is necessary if trends are to be discovered from surveillance data. In infectious disease in particular, temporal trends are helpful. In predicting and planning for disease there may be a significant delay between the time of acute illness and the date the infection is diagnosed and then reported. This is generally unavoidable. First, the time of onset of symptoms may be days or weeks before the illness is investigated. There will then be further delays before the disease is diagnosed, and again a delay before it is reported to the surveillance unit. With serological diagnosis, because a rise in antibody occurs only during convalescence, the infection is only confirmed when the patient is better. Even with a weekly reporting system there will be further delay before the infection is finally recorded. The burden of reporting will be increased considerably by asking reporting laboratories to record the date of onset of illness, but the date the first specimen was received in the laboratory is usually readily available and often approximates to the time of acute illness. In the analysis of such data, the interval between this date and the date of reporting may need to be taken into account. Analysing laboratory data by date of first specimen may be less helpful for the immediate detection of changes in incidence. The techniques of analysis by time, and tests for seasonality or periodicity, are outside the scope of this chapter.
The site of the reporting laboratory is normally taken to be the geographical location of the ill person. Except in the best organized of health-care systems, however, this is not always true. In England and Wales, for example, the statutory notifications generally relate to the area of residence of the patient, but the Local Authority to which the infection is notified is not necessarily coterminous with the Health Authority. With laboratory data there are rarely rigid boundaries or catchment areas for hospital or public health laboratories. It may be difficult or impossible to allow for these problems in the analysis of the data, although it may only rarely be necessary to do so.
Analysis by place may show different patterns of infection such as north–south or urban–rural. Generally, national surveillance schemes will point only to broad differences in incidence; for small area differences surveys are usually necessary, except with more sophisticated surveillance systems. Analysis by place is particularly important in outbreak investigations.
Age and sex of cases reported are also important components of a surveillance report. It is possible to conduct a surveillance system based solely on numbers and geographical location, but the sensitivity of the systems will be considerably reduced. Surveillance based on laboratory-reported rubella in England and Wales has shown that, although the incidence in all age groups fell substantially following the introduction of the measles, mumps, and rubella (MMR) vaccine in 1989, two epidemics occurred subsequently, one in 1993 and then in 1996 (Fig. 1). However, it was quite clear from surveillance data that the main force of the epidemic fell on adults, especially young adult unvaccinated males (Fig. 2) (Miller et al. 1997). Few pregnant women contracted the infection. In an outbreak of hepatitis B caused by a tattooist in one district of England (Limentani et al. 1979), a simple analysis of notifications of ‘infective jaundice’ by time or place would not have uncovered the outbreak. However, analysis of notifications by age and sex revealed an increase in notified cases in males aged 15 to 29 years.
Fig. 1 Rubella: laboratory reports 1984–1996, England and Wales.
Fig. 2 Rubella: sex and age distribution, laboratory reports 1993 and 1996.
In the laboratory surveillance system in England and Wales, some important and common organisms, such as salmonellas, were reported without the age and sex of the patient, as it would have increased the burden of reporting, collection, and analysis of such data in a pre-computer age to unmanageable levels. Reliance was placed on the ability to detect an increase in a rare Salmonella serotype, or the phage type of a more common serotype. When this occurred, the age and sex of the cases were readily available from the reference laboratory, and often provided some aetiological clues. This relatively crude and insensitive system nevertheless proved valuable, as when an increase in rare Salmonella ealing infections proved to be mainly in infants and was subsequently proven to be caused by powdered milk (Rowe et al. 1987). In this outbreak the epidemiological evidence was extremely strong while the level of contamination of the milk by the Salmonella was remarkably low.
Denominator data clearly are desirable, but they are not essential to an infectious disease surveillance system. In the United Kingdom only the Royal College of General Practitioners’ (RCGP) Research Unit has an in-built rate provided with its data outputs. In this surveillance system, about 40 physicians in various parts of the United Kingdom, covering about 200000 patients, report new episodes of illness each week to the Central Research Unit (Fleming and Crombie 1985). As each GP has an age and sex distribution available of his practice populations, consultation rates for the different illnesses can be calculated.
In the interpretation of surveillance data lies the skill of the epidemiologist. As Langmuir (1976) wrote of William Farr:
His weekly return was no archive for stale data, but with his facile pen became a literate weapon for effecting change. He presented his analysis with objectivity but then stated his own interpretations forcefully…
In surveillance data a detailed understanding of the reporting system is necessary before a meaningful interpretation of the statistics can be made. Some knowledge of the size and demographic characteristics of the population covered by the surveillance system is also necessary. Every data source carries it own strengths and biases (Moro and McCormick 1988). With each data source, timeliness, completeness, representativeness, and accuracy (Thacker et al. 1983) should be considered. To these four qualities should be added that of significance (in its sense of ‘importance’). ‘Timeliness’ may be particularly important with laboratory statistics, where there may sometimes be a considerable delay between disease onset and laboratory diagnosis. Organisms need time to grow in culture, and acute and convalescent samples of sera are needed to demonstrate the fourfold or greater rise in antibody necessary to substantiate a serological diagnosis. Rapid diagnostic methods for identification of an organism and the increasing use of IgM tests, however, have hastened considerably the diagnostic process for some infections. Notifications, although usually made on the basis of a clinical diagnosis, may not always be as prompt as one would expect, and sentinel GP clinical reporting systems tend to be more timely (Tillett and Spencer 1982).
The need for ‘completeness’ of reporting, particularly of common infections, is often exaggerated. Detection of disease trends by time, place, and person, sufficient for meaningful epidemiological interpretation, is easily feasible with incomplete data. Indeed, striving for completeness may waste resources. For rare diseases, however, or diseases that have become rare following a control programme, completeness grows in importance. Passive surveillance systems rarely achieve completeness; active systems are generally necessary for this.
The ‘representativeness’ of the data collected needs thought and planning, and the advantages of collecting data from more than one source may provide ways of validating one data source against another. In the surveillance system for infectious diseases in England and Wales conducted by the Communicable Disease Surveillance Centre, for example, the three main sources of data on meningococcal meningitis—hospital, notifications, and laboratory—show similar trends over time (Fig. 3). Source data should be representative for time, place, and person.
Fig. 3 Acute meningococcal meningitis: three sources of data, England and Wales, 1957–1988.
The ‘significance’ of surveillance data should also be evaluated carefully in its interpretation, and this can perhaps be best illustrated diagrammatically, using laboratory reporting of influenza as an example.
If a circumscribed population affected by an outbreak of influenza is represented by the rectangular outline in Fig. 4, the number actually infected will be a proportion of this, represented by the circle (A). However, not all of those infected will have symptoms; those that do belong within circle (B). Only a proportion of these will visit a doctor (C) (whether in hospital or general practice), and progressively smaller proportions will have a specimen sent for examination in a laboratory (D), specimens positive (E), and positive specimens reported to the surveillance unit (F). The biases and variables that occur during these steps also need to be considered. Those with symptomatic infections may be those never previously exposed to the particular influenza variant or subtype, or the very young and the very old, or those with a chronic disability, such as a respiratory condition. Similarly, those who visit a doctor and those whom the doctor investigates may be influenced by several factors, including social class, age, and severity of disease, and proximity to a laboratory. Laboratory success in its turn will depend on availability and cost of reagents, the interest and expertise of the laboratory, and the age and severity of disease in the patient. Finally, the accuracy, completeness, and timeliness of reporting by the laboratory will be influenced by its motivation, organization, and efficiency, as well as by the usefulness of the surveillance system and the quality and value of its feedback. A similar progression can be worked out for notification, GP, hospital, and death certification data sources (Fig. 5).
Fig. 4 Stages in the reporting of laboratory infection.
Fig. 5 Stages in the progression of disease.
The ‘accuracy’ of the data provided by the laboratory will depend not only on its interest, expertise, and motivation, but also on the clarity of the instructions for reporting provided by the surveillance unit. These include acceptable diagnostic criteria for laboratory data (especially for antibody titre measurements, the levels acceptable to the collection unit should be clarified) and clinical definition for notification and GP data. A Quality Control scheme for participating laboratories is useful, as in the system run by the Public Health Laboratory Service in England and Wales.
Dissemination of information and target groups
The logical end, and indeed the purpose of any surveillance system, is the output, and its quality and relevance are critical: not only must it be meaningful and intelligible, but it must also be directed at the appropriate targets, whether they be decision makers or research workers. Moreover, feedback to those who provide the reports has an important motivating role in any surveillance system. Some surveillance and monitoring systems may even improve performance, as in the COVER (cover of vaccination evaluation rapidly) programme in England and Wales. In this surveillance and monitoring system the measurement and publication of performance measures for vaccine coverage in different health districts almost certainly stimulates poorly performing districts to improve (Begg et al. 1989).
For infectious disease surveillance systems, weekly or monthly reports are the most appropriate. The ability to adapt to changing patterns of infection—incorporating new diseases of importance, discarding outdated or useless data needs to be in-built. Flexibility to provide urgent information is useful, as for outbreaks, and as methods of communication continue to improve this should make urgent dissemination easier.
The rapid development and availability of electronic methods of storing, analysing, and disseminating data and information have greatly facilitated the practice of surveillance. Powerful microcomputers and electronic networking have allowed data to be transmitted from source to surveillance centre without the use of post and paper—or indeed the tedious completion of surveillance forms. Not only can data be stored in the computers, but pre-programmed statistical analyses of the data can also reveal changes in time, place, or person. Fortunately, the meaningful interpretation of such changes still requires the human element. Finally, the dissemination of information has been revolutionized by the use of electronic methods. In the United Kingdom, the Communicable Disease Surveillance Centre provides rapid information to Health Authorities through its EPINET System. In France, the Minitel System is also effective.
Content and presentation
The summary report should not contain indigestible lists and tables, but easily understood analyses with appropriate evaluation of their significance. Especially for infectious disease, the reports need to be topical and relevant. Thus short summaries of recent trends and changes, together with more detailed reviews of subjects and interests, are important ingredients of a surveillance report, and these can be supplemented by reports of outbreaks and other items of general interest by reporter participants or other contributors.
A successful report will educate and provide current scientific information for planning, prevention or change. (Eylenbosch and Noah 1988).
Another function necessary for a successful surveillance system is the provision of an information service for individual enquiry. The organization of the surveillance unit and of the data collection system to provide information (as opposed to raw data) is as important a part of any surveillance feedback service as the regular report. For the sporadic inquiry of this type, appropriate interpretation of the information provided is also necessary.
Encouraging regular dialogue between providers, other interested parties, and the central surveillance unit is helpful to fostering a healthy relationship between them and the long-term usefulness of the surveillance system; it can also be regarded as a form of monitoring of the surveillance system.
Ideally, the content and presentation of the output of the surveillance system needs to be adapted so as to be made intelligible to each type of target group; lay politicians and decision makers, for example, might receive a different type of feedback from research workers, and again from the public and media. This is rarely possible with the regular surveillance report, but an information section within the central unit could tailor the response appropriately to the ad hoc inquiry. The increasing interest in health by the lay public, media, and politicians makes it essential to provide accurate, relevant, and topical information with skill and flair.
Content of surveillance: sources of data
When discussing the epidemiology of infection, or any other type of disease, it is useful to consider the different stages in the natural history of the disease process (Fig. 5). The figure is similar to that used for laboratory data (Fig. 4). The population again can be represented by the rectangular outline.
Within this population will be a subpopulation who will be infected asymptomatically, or will be immune or carriers. A smaller proportion will develop symptomatic infection, and progressively smaller proportions will have symptomatic unreported infections, visit a doctor, be admitted to hospital, or die. To have a true measure of the total impact of a disease on a population, information at all these levels is needed. In practice it is of course rarely possible, or perhaps necessary, to be so systematic, although information at most levels can often be obtained. Serological surveys will give information on asymptomatic infection or immunity levels, and the taking of appropriate specimens or swabs from persons may detect asymptomatic carriers. Surveillance of predisposition to disease other than that determined by the absence of antibody is more difficult in the field of infectious disease, but surveillance of general functions such as growth, development, and nutritional status of children (Morley 1975; Irwig 1976; Carne 1984) may fall into this category, as will surveillance for infection in certain groups, such as tuberculosis in certain ethnic patients, human immunodeficiency virus (HIV) infection in haemophiliacs and homosexuals, cytomegalovirus infections in the immunosuppressed, or in those subject to certain procedures such as urinary catheterization. Unreported morbidity (Fig. 5) is not usually possible to place under passive surveillance. In a series of surveys conducted by the Office of Population Censuses and Surveys of England and Wales, the General Household Survey, information on unreported morbidity is obtained (Haskey and Birch 1985). Although each survey is finite, the collective information over many surveys constitutes a database suitable for surveillance.
A general practice surveillance system based on a sentinel reporting network was first successfully organized in the United Kingdom by the RCGP in 1966 (Fleming and Crombie 1985), and since then in the Netherlands (Collette 1982) and Belgium (Thiers et al. 1979). General practice morbidity data are generally useful for providing information one tier in severity below that of hospital morbidity (Fig. 5). More specifically, in practice they produce information on clinical conditions with very low mortality not covered by notification, such as the common cold, chickenpox, or otitis media. In England and Wales, before mumps and rubella became notifiable in 1988, the RCGP clinical reporting system was an important source of information on these common infections; it remains an important source of information on clinical influenza, which is not notifiable. Sentinel reporting systems are generally inappropriate for rare diseases. When, as in the United Kingdom, the reporting GPs keep an age–sex register, a more accurate denominator is available than that used for notification, and indeed the RCGP data are published as rates. Most countries have a notification system that usually provides an essential source of information on the important communicable diseases; the characteristics of these systems are well known and will not be discussed in detail here. Notifications cover both general practice and hospital morbidity. It is often more useful to know what the notification rate is than to strive for completeness, which is generally only really desirable for very rare or very serious infections. Diseases that are perceived by the notifying doctors to be either a serious public health problem and communicable are often better notified than those that are not. Thus tetanus is often poorly notified in countries such as England and Wales. It must be remembered that the primary objective of a notification system is not for surveillance but to provide an opportunity for local control, for which legal powers are usually available if necessary.
Surveillance systems using hospital admission data clearly only cover a limited, although important, phase in the natural history of an infection. Infections for which hospital data are ideal are those for which patients are usually admitted and admitted once only, and in which the diagnosis can be confirmed. If the hospital reporting system is based on a sample, the disease should not be excessively rare. Meningococcal and other forms of bacterial meningitis are examples of infection that tend to be well documented in hospital data systems. Hospital data, on the other hand, may be available late, months or even years after the event.
Death certification is virtually a universal requirement in most countries. Death certificates are clearly important to any surveillance system, but for infectious diseases their value may be somewhat limited as infections are now less common as a primary cause of death. As infections remain an important cause of morbidity, death certifications need to be supplemented by one or more of the surveillance systems detailed above. They also tend to be too imprecise in deaths attributed to infection.
Other sources of data
Laboratories provide important, perhaps essential, information for the surveillance of infection. Surveillance, supported by laboratory confirmation of clinical cases reported, showed that the malaria epidemics said to be occurring in the malarial southern states of the United States did not exist, and the few cases confirmed microbiologically were imported or relapses (Langmuir 1963). In addition to this ‘confirmatory’ role in surveillance that laboratories provide, they have an additional ‘qualitative’ feature. Thus laboratory data cannot only confirm the presence or absence of influenza, but their data can also show whether the virus is type A or type B, what the subtypes or variants are, and whether these have changed since the previous epidemic (Fig. 6). Similarly, Mycoplasma pneumoniae, psittacosis, Q fever, several viruses, legionella, and many other agents may cause atypical pneumonia, and only a laboratory can distinguish these successfully. In human infections, laboratories can provide data at all levels of the disease process shown in Fig. 5. Laboratory data often provide information on infection in vectors, animal, other hosts, or the environment, and are especially important with zoonoses.
Fig. 6 Influenza surveillance, England and Wales. Indices used in monitoring influenza activity for one major epidemic winter (1975–1976) are compared with five other winters. (Source: Communicable Disease Report, 44, 3–4, 1989.)
Surveillance of laboratory-diagnosed infections can in itself lead to early recognition of outbreaks. Several examples of this have been reported, especially with salmonellas (Gill et al. 1983; Rowe et al. 1987; Cowden et al. 1989), and with legionellas (Salmon and Bartlett 1995). Surveillance of outbreaks can also provide useful information, especially in assessing trends, and may be particularly worthwhile in countries where more sophisticated sources of data, such as laboratory data on individual infections, may not be available. Outbreak surveillance can be cheap and effective, although it is essentially an insensitive measure of disease trends.
Surveillance of vaccine utilization is an important component of the process of monitoring the effect of vaccine strategy on an infection. In England and Wales the COVER programme fulfils this function (Begg et al. 1989) and may have an additional effect in stimulating poorly performing districts to improve coverage. Serological surveillance can provide an indirect measure of the effectiveness of a vaccine strategy (Noah and Fowle 1988).
Sickness absence records may provide indications of major outbreaks in working populations. Influenza in particular may produce measurable changes in sickness absence. Sickness absence can also be monitored in special groups, such as in boarding schools, and among Post Office workers. Sickness absence is a crude measure of a widespread epidemic, and covers only a selected age group of the population, but can be surprisingly sensitive and can provide an early warning of an epidemic. Only a limited number of countries have this type of information available.
Biological changes in agent, vector, and the reservoirs of infection can be placed under surveillance. Surveillance of changes in an agent, such as new subtypes or variants of influenza, antibiotic resistance in bacteria, such as the gonococcus and Staphylococcus, or protozoa, such as Plasmodium, is regularly performed. Surveillance of biological vectors such as ticks and mosquitoes, and of animal reservoirs of infections such as brucellosis and rabies, is an essential component of disease control in many countries.
Susceptibility to infection can be measured by skin testing or serological surveillance. In England and Wales, antibody profiles to current circulating influenza variants and to new variants are regularly performed in small samples of the population to assess the degree of susceptibility to a new variant. The degree of immunity to vaccinatable diseases is also regularly assessed (Morgan-Capner et al. 1988). Raska (1971) persuasively argued the use of serum banks and immunological surveys in surveillance.
Other conditions that can be placed under surveillance include abortions, birth defects, injuries, behavioural risk factors, and occupational safety.
Objectives of surveillance
Many of the objectives of surveillance of infectious diseases have already been alluded to in the text and will be summarized here only. The main objectives of infectious disease surveillance are to monitor disease trends in time, place, and person, as illustrated below.
Anticipating changes in incidence
Many infectious diseases follow regular patterns, both seasonal and secular. The respiratory syncytial virus follows a distinct seasonal pattern, causing epidemics every year with the peak incidence, in the Northern Hemisphere, almost invariably shortly after the new year (Fig. 7). Minor variations in this pattern occur (Noah 1989). With some viruses, for example, echovirus, a failure to return to baseline by the end of its yearly cycle signifies that a resurgence will occur the following year (Fig 8 and Fig 9) (Epidemiology Research Laboratory 1975). Some organisms, such as Mycoplasma pneumoniae, have long cycles extending over 4 years, but are particularly important as this is a treatable infection and early warning of an epidemic is of practical value (Fig. 10) (Noah 1974; Monto 1974). A review of cyclic variation in infections can be found in Noah (1989).
Fig. 7 Laboratory reports of respiratory syncytial virus infections, England and Wales.
Fig. 8 Laboratory reports of echovirus 19 virus infections, England and Wales.
Fig. 9 Laboratory reports of echovirus 4 infections, England and Wales.
Fig. 10 Laboratory reports of M. pneumoniae infections, England and Wales.
Early detection of outbreaks
Outbreaks of food poisoning which have been detected only because surveillance revealed an unexpected increase in a salmonella have already been described, and reports of some of these have been published (Gill et al. 1983; Rowe et al. 1987; Cowden et al. 1989). Occasionally, early detection of a new strain of an organism may lead to premature action which, with hindsight, turns out to have been inappropriate, as with the swine influenza episode (Silverstein 1981). Surveillance may lead to the detection of new infections (e.g. Lyme disease).
In an outbreak of hepatitis caused by a tattooist in England in 1978, surveillance by time alone would not have brought the outbreak to light, but only the characteristic age and sex distribution (young adult males) showed a change (Limentani et al. 1979). Analysis of surveillance data by person, time, and place is more likely to reveal changes in pattern than analysis by one of these parameters alone.
Evaluation of effectiveness of preventive measures
Surveillance techniques are used to monitor the effects of a mass vaccination programme. Not only can changes in incidence be measured by time (Hinman et al. 1980), but also by person; for example, age changes in measles or mumps (Hinman et al. 1980; Cochi et al. 1988) have been recorded as a result of mass immunization. There have also been examples of monitoring changes in incidence by place (district) and correlating these changes with vaccination uptake rates (Pollard 1980). Effectiveness of a vaccination programme can also be assessed by serological surveillance (Noah and Fowle 1988). The effect of withdrawing a source of infection, such as a contaminated food, from circulation, can be monitored by surveillance, as with the Salmonella napoli outbreak caused by contaminated imported chocolate in England (Gill et al. 1983; Fig. 11).
Fig. 11 S. napoli outbreak in England and Wales, 1982: distribution of 202 primary household cases.
Identification of vulnerable groups
Surveillance can expose vulnerable groups; for example, by revealing ethnic or social differences in tuberculosis incidence. Appropriate action, such as BCG vaccination of neonates in such groups, can be taken. Serological surveillance can also identify susceptibility in particular groups of persons for selective vaccination.
Setting priorities for allocation of resources
From the examples above, it is clear that surveillance programmes can be used to provide information for setting priorities for resource allocation, and hence for planning. This leads to more efficient use of health resources and, although changes cannot always be anticipated in advance, prompt detection of changes can allow redistribution of resources at an early stage. Surveillance can also be useful in uncovering or monitoring changes in health practices, e.g. the increasing rate of Caesarean section in the United States between 1970 and 1988 (Thacker 1994).
Infectious disease patterns may help to generate hypotheses for aetiology of chronic diseases. Secular variation compatible with certain infections have been noted with sudden infant death syndrome (Helweg-Larsen et al. 1985), insulin-dependent diabetes (Gleason et al. 1982), deaths from asthma (Khot and Burn 1984), and anencephalus and spina bifida (Maclean and Macleod 1984).
Setting up a surveillance system
In setting up a surveillance system, all the points discussed above need to be considered. It is important in addition to ‘pilot’ the system first, or to have a ‘trial period’ after which necessary adjustments can be made. In practice, as surveillance is essentially an ongoing process, adjustments may be made continually to ‘fine tune’ the system. Unnecessary extraneous or duplicate reporting should clearly be avoided, and improvements will also need to be made in the methods of flow of the information. Three attributes mentioned by Klaucke (1994) for an efficient surveillance system are simplicity, flexibility, and acceptability. Improving methods for the flow of information to and from the surveillance unit will enhance these three attributes. Electronic methods of reporting, obviating the use of paper, will for example not only simplify the system and reduce the effort involved, but will also increase its acceptability. Moreover, electronic systems make it easier for flexibility to be built-in, so that conditions under surveillance can be added or deleted more easily. The surveillance system should be considered as a network of roads along which the types of consumer product sent can be changed according to need. In England and Wales, when rubella and mumps vaccines were introduced into the routine schedules, the two infections were placed under surveillance using the existing notification network. The laboratory network was used for AIDS, legionnaires’ disease, and Lyme disease. Thus, surveillance systems for new conditions were put into operation rapidly, fairly painlessly, and at minimal extra cost.
Acceptability depends on simplicity, and also on the perception by reporters of the importance of the data being collected. Quality of feedback, both in the initial stage of setting-up the system and throughout the operation of the surveillance, clearly affects acceptability.
The technology now available to facilitate surveillance is considerable. Electronic reporting in both directions—reporter to surveillance unit, surveillance unit to reporter and others—is now feasible. Details of the computer technology available are outside the scope of this chapter, and in any case are rapidly changing.
It is now becoming increasingly obvious that ‘the links between seemingly far-flung events involving pathogens and their hosts deserve a new kind of scrutiny if we are to deal effectively with emerging and re-emerging infectious diseases’ (Fox 1996). The scope of surveillance must increase indefinitely in the next few years so that the effects on infection of changing environments (whether natural or man-made) can be anticipated—and monitored. We need to look not only on human health but on effects on flora and fauna also, because they each affect each other in ‘delicate balanced fragility’.
Although considered to be the backbone of public health, surveillance makes use of data gathered for other purposes. Clinical or laboratory diagnoses are generally made for the benefit of the doctor and patient. Thus, surveillance is an efficient way of making use of data that has in effect, ‘passed its sell-by date’. ‘…probably most important, surveillance needs to be used more consistently and thoughtfully by policymakers’ (Thacker 1994). It is up to public health physicians and epidemiologists to transform routine statistics into meaningful reliable and timely information so that health policy-makers can be persuaded to act for change—for change is the ultimate goal of surveillance.
Field investigations of vaccines
Vaccines are a time-tested and highly efficient means of controlling many microbial threats. Epidemiological studies are an essential part of the assessment and use of vaccines and may be associated with vaccines in several different ways: (a) in the assessment of the need for a vaccine by undertaking surveillance or surveys; (b) in antibody and field trials to assess the efficacy and safety of the vaccine; (c) in the selection of an appropriate strategy; (d) in the implementation of the vaccine programme. Finally, the use of the vaccine and its effect on the population need to be closely monitored. These logical steps have not always been followed.
Assessment of the need for a vaccine
Morbidity and mortality of disease
The overall morbidity of the infection, and its severity, can (and should normally) be estimated from a surveillance programme. With measles the need for a preventive programme in the United Kingdom was first shown in 1963 (McCarthy and Taylor Robinson 1963). Miller (1964), by studying notifications of measles, was able to estimate the hospital admission rate (1 per cent), the respiratory and otitis media complication rate (6 to 9 per cent), and the neurological complication rate (0.7 per cent). Complication rates like these for an extremely common infection signified a serious health problem. Miller (1978) showed that, although the incidence rates had fallen with the introduction of mass vaccination, these complication rates had changed little with time. Rey (1985) estimated that in France every year 30 people died from measles, 6000 were admitted to hospital, 100 to 200 developed encephalitis, and 10 to 20 subacute sclerosing panencephalitis. Worldwide estimates tend to be cruder but none the less possible; using world health statistics, Cutting (1983) claimed that measles caused 1 to 1.5 million deaths in children in 1981. General practice surveillance systems may also be useful in providing estimates of disease burdens, as for mumps in the United Kingdom (Research Unit of the RCGP 1974). Serological studies are also used to assess the overall impact of an infection on a population, and the vulnerability of a target population. Studies on rubella in 1969 (Cockburn 1969) showed that by adulthood more than 80 per cent of a population had been infected leaving 5 to 15 per cent of pregnant women still susceptible and hence, still vulnerable. In the United States the incidence of clinical rubella in pregnant women was found to vary from 4 to 8 per 10 000 in endemic periods to 20 per 10 000 during epidemics, while subclinical infection could increase by up to threefold (Sever et al. 1969; White et al. 1969). Sometimes, as in poliomyelitis, crude notification data on the most serious outcomes of poliovirus infection—paralysis and death—can be sufficient to point to a need for a vaccine (Fig. 12). For severe and very rare diseases only a selective vaccine policy can be considered, so these considerations do not apply.
Fig. 12 Poliomyelitis notifications: England and Wales 1919–1983. (Prepared by the Communicable Disease Surveillance Centre.)
Field studies of vaccine efficacy (VE) and safety are clearly necessary before a vaccine is licensed for use. Postlicensing studies are also important, but these require a different approach and will be considered separately.
Pre-licensing vaccine trials
The earliest trial (phase 1) of a new vaccine usually involves a small number (usually 20 to 50) of adult volunteers with a low level of risk of acquiring the disease. Antibody is measured and adverse effects are closely monitored. This tests the immunogenicity of the vaccine, following which the dose may be adjusted. Preliminary information on safety is also provided by a phase 1 trial—although limited, this is an important aspect of such a trial.
In the phase 2 trial a larger number of persons, between 100 and 200, constituting a target population is vaccinated. Their risk of acquiring the disease, again, should be low. Efficacy is again estimated by serological testing, and more precise information on dose–response and safety obtained, with some estimates of the rates of the more common side-effects. The number of doses necessary and some preliminary information on contraindications may also become known after a phase 2 trial.
If the phase 2 trial is satisfactory, a large field trial, the phase 3 trial, of perhaps more than 500 high-risk subjects can begin. Vaccine protection is tested against disease acquisition. Questions that need to be answered in a trial like this will include the efficacy of the vaccine, both in degree and duration, the rate of side-effects, both rare and common, the optimum age for giving vaccine, the dose, including how many and at what intervals, and the need for any adjuvants. Further contraindications to the vaccine may become apparent.
VE can be assessed by measuring disease incidence, by serological test, or both. Important considerations in trials based on disease incidence are how much the two groups (case and control) vary in exposure to the infection, and how infection is ascertained in each group.
If, as is generally believed, vaccinated groups tend to consist of those who are of higher social class, and are thus groups which make better use of health services, diseases such as tuberculosis or whooping cough may be less common in them, or less severe, and hence, less likely to be ascertained. While any such disease is more likely to be reported, these confounding factors are unlikely to cancel out each other. Allocation of persons to vaccine or control groups must be random and double blind to minimize the effect of such biases. This is generally known as random individual allocation. Group allocation—class, village, factory—can also be used in vaccine trials; although the control group must also be matched carefully, it is not usually possible to allocate randomly or for the trial to be double blind. Detailed descriptions of how to conduct trials by individual or group allocation are given by Pollock (1966).
The disadvantages of a randomized controlled clinical trial are those of any longitudinal or cohort study: the necessity for large numbers of patients, the expense, the high drop-out rates that occur with long follow-up, and the fact that the vaccine is used under ideal but artificial conditions, the epidemiological equivalent of in vitro.
These limitations make postlicensing vaccine trials important (sometimes known as phase 4). Continued assessment, or monitoring, of vaccine use after licensing is mandatory (the equivalent of in vivo). Vaccine potency may change, either because of some change at the manufacturing level, or along the chain that gets it to the user, or in storage by the user or as a ‘natural’ effect with time. The live varicella-zoster vaccine is manufactured to contain 3000 pfu/dose. This is calculated to reach the minimum dose of 1350 pfu by the expiry date, assuming storage under recommended conditions (Arvin 1997). Thus monitoring of vaccines at all levels of the chain is important, especially as vaccines are often used in different climates and different population groups. Rare side-effects may become apparent. In postlicensing trials the populations immunized may be more or less responsive than in the pre-licensing trials on account of a difference in age, social class, or other factors. For these reasons, complacency is unwise after a successful large-scale pre-licensing trial.
The success of any vaccine programme will depend on the overall population immunity level achieved. This is a function of efficacy and uptake of vaccine (Noah 1983):
vaccine efficacy × vaccine uptake = population immunity.
The overall immunity level target required will depend on the infectiousness of the disease and on the goals of the vaccine programme, whether it be containment, elimination, or eradication.
VE is at its optimum when the vaccine leaves the manufacturer: factors that may influence efficacy after this include the age of the patient, the site of the injection, the immunological status of the patient, the conditions under which the vaccine has been transported, stored, and made up. In addition, for unknown reasons, the vaccine may not protect some people. Factors influencing the assessment of VE are considered in more detail below.
Vaccine uptake is primarily influenced by the efficiency of the administrative processes in a country or province, but can also be affected by public perception of the vaccine itself (which may be rational or irrational) and of the disease. Generally, a well-organized system for mass vaccination is much more efficient than sporadic vaccination campaigns. Thus, postlicensing studies include not only continuous monitoring of VE, but also of uptake and implementation of the vaccine.
Evaluation of outcome and of side-effects can also be studied.
Assessment of vaccine efficacy
Assessment by incidence
The incidence of the disease in a population after vaccination is compared with that before vaccination. This is a simple and fairly universal undertaking, and is a useful method of assessing the impact of a mass vaccine programme on a population. VE cannot usually be calculated from this method, only the effectiveness of the vaccine programme in broad terms. Moreover, demonstrations of a reduction in incidence following vaccination shows an association that may not be causal, and the well-known rules of Bradford Hill (1971) for assessing causality from association apply. Changing social circumstances, changes in the population, or a natural variation in the incidence of the disease may have caused the change in incidence. In England and Wales scarlet fever declined in both incidence and severity without a vaccine. The use, first, of Salk polio vaccine in 1957, followed by the Sabin vaccine in 1963, reduced the incidence of poliomyelitis to negligible numbers (Fig. 12); however, the introduction of BCG in 1950 had a less than dramatic effect on an already declining disease (Fig. 13). This does not necessarily mean that the BCG vaccine was worthless: a reduction in incidence could have been masked by more complete ascertainment or better diagnosis of cases, and accompanying increase in transmission of or susceptibility to tuberculosis, or a selective decrease in morbidity from a severe but rare form of tuberculosis (e.g. miliary tuberculosis or tuberculosis meningitis), too small to be detected by crude methods. Detailed surveillance would be necessary to detect such changes.
Fig. 13 Respiratory tuberculosis: England and Wales 1913–1988. (Prepared by the Communicable Disease Surveillance Centre.)
Assessment by immunological testing
Immunological testing includes skin tests, such as tuberculin or Schick, and serological testing. Tuberculin testing, generally using the Heaf or Mantoux tests, is performed to assess a person’s immunity before vaccination. It has also been used to assess the quality of different strains of BCG. Schick tests for diphtheria are now performed rarely. The local reaction to smallpox vaccine was used to assess whether or not the vaccine had ‘taken’.
Serological testing can be performed using antitoxin (diphtheria and tetanus) or antibody (measles, rubella, etc.) levels. Serological tests first have to be shown to correlate with immunity to disease by disease incidence studies. Studies using such tests can be of two types: seroconversion or seroprevalence (Orenstein et al. 1985).
Seroconversion Seroconversion studies are generally used in phases 1 and 2, and sometimes also in phase 3 of pre-licensing trials. They are particularly useful for assessing vaccine response at different ages, as with the measles vaccine, and in different groups of persons; for example, healthy persons and those with immunosuppressive conditions respond differently to pneumococcal vaccine. Seroconversion studies are also useful for assessing VE for rare conditions in which antibody levels have already been shown to correlate with protection, as with tetanus, diphtheria, and poliomyelitis. Moreover, seroconversion studies have the advantage that any change in antibody status can be attributed to the vaccine. On the other hand, seroconversion studies require two blood samples and laboratory back up, but these disadvantages can be weighed against the smaller number of subjects that need to be studied. A more serious disadvantage is the absence of a reliable serological test of immunity for some infections, such as whooping cough and meningococcal infections.
Seroconversion studies and incidence studies suffer from the same problems of having a case definition that is too specific (too high an antibody level or too rigorous criteria for diagnosis) thus underestimating efficacy, or one that is too sensitive (too low an antibody level or too loose criteria for diagnosis), which will overestimate efficacy.
Seroprevalence Seroprevalence studies have the advantage of requiring only one blood test but the corresponding disadvantage is relating a ‘positive’ result to disease or vaccine. Seroprevalence studies can be used to monitor success of the vaccination programme by measuring overall population immunity, which is a function of both efficacy and uptake. Thus, the prevalence of poliovirus antibody in a highly immunized population with no wild virus disease will reflect the efficacy and uptake of the vaccine. The prevalence of rubella antibody in antenatal women or women of childbearing age in a country with a selective rubella vaccine policy (so that the natural epidemics of rubella are virtually unaffected) will measure the overall immunity status of the target group, but it will not easily distinguish how much of this is attributable to natural infection and how much to the success of the vaccine programme (Noah and Fowle 1988). If the expected seroprevalence of rubella antibody in this target group under conditions of natural rubella endemicity is known, then the gain in immunity attributable to vaccination of the target group can be calculated.
With seroprevalence studies the duration of protection can also be evaluated, but Marks et al. (1982) pointed out the importance of the timing of immunity measurements since vaccination, and also the age at vaccination. In case–control studies using seroprevalence, these two parameters must be similar in cases and controls; that is, the time lapse between vaccination and testing immunity and the age at vaccination are crucial for matching.
In seroprevalence studies the same caveats apply about sensitivity and specificity of the antitoxin or antibody levels chosen to indicate immunity. In seroprevalence studies the choice of which class of antibody (IgM or IgG) to measure can also be critical.
Seroprevalence studies related to previous vaccination require accurate records of previous vaccination (including timing and number of previous doses of vaccine) for meaningful interpretation (Orenstein et al. 1988). Absence of antibody may not always correlate with susceptibility.
Assessment by epidemiological studies
For epidemiological studies of VE after licensing, a number of general comments need to be made. As with seroprevalence studies, the sensitivity and specificity of the case definition is important, and has always to be a compromise. A highly sensitive case definition will include many spurious cases, leading to a falsely low VE, whereas too specific a case definition will lead to a falsely high VE. Case definition in both control and vaccinated groups should be undertaken with equal vigour, with due awareness of the tendency for case definition to be more specific in the vaccinated group. Exposure in the two groups should be similar; even if vaccinated and control groups have been closely matched by age, sex, social class, and geographical distribution, variations in exposure may still occur. The diagnosis should if possible be made ‘blind’ (Marks et al. 1982). In retrospective studies, ascertainment of vaccination history in the two groups must also be pursued with equal vigour, and again should preferably be done ‘blind’. Those with a previous history of the infection should not be included in either group of subjects.
An outbreak often affords an opportunity to assess VE in a field setting (in vivo). The basic steps are to define the cohort under investigation, ascertain all cases according to an acceptable case definition, ascertain vaccination histories in the entire cohort, and calculate attack rates in vaccinated and unvaccinated persons, excluding those with a previous history of disease. For very large cohorts sampling or cluster sampling (Henderson and Sundaresan 1982) can be used. In infections in which most cases occur in fairly well demarcated age groups, the cohort under study will usually exclude those outside these age groups, as well as those too young to have been vaccinated. The optimum age of vaccination can also be ascertained from outbreak studies (Judelsohn et al. 1980).
Secondary attack rates in households
Measuring the attack rates in members of a household (secondary cases) following the introduction of an infection by one member of the family (primary case) affords an attractive approach to VE studies because exposures in vaccinated and unvaccinated persons are similar, thus eliminating an important bias in these studies, and also because the denominators (those exposed) can be fairly accurately determined. These types of study, however, need more careful planning than most other postlicensing studies, and they are usually better designed as prospective studies. They also need to be planned to take place during periods of high epidemicity of the infection. Studies on pertussis vaccine in England and Wales, which used secondary attack rates in households, suggested that the vaccines were ineffective (PHLS Whooping Cough Committee and Working Party 1973; PHLS Epidemiological Research Laboratory 1982). This was difficult to believe because the incidence of whooping cough in the country at the time of the first study had been decreasing steadily, and other studies (Noah 1976; Pollard 1980; PHLS Epidemiological Research Laboratory 1982) suggested that the vaccine was effective. The reasons for the failure of the household exposure studies to confirm efficacy were examined by Fine et al. (1988); they attributed the apparent lack of VE primarily to the inclusion of retrospectively ascertained cases and of households in which the primary cases constituted a vaccine failure. Secondary attack rates in households have also been used to measure the efficacy of hepatitis A vaccine in preventing secondary cases of the infection (Sagliocca et al. 1999)
Cluster sampling (Henderson and Sundaresan 1982) is a modified form of this method of estimating VE. It has the advantage of being cheaper and easier to conduct than a household study, and hence particularly suitable for developing countries, but it is also less rigorous (Orenstein et al. 1985).
Retrospective population cohort studies
This type of study depends on the availability of fairly sophisticated disease reporting and vaccination recording systems. For each case of disease ascertained, the vaccination history is verified, and from the number and percentage of those vaccinated in the base population, the attack rates in those vaccinated and unvaccinated can be calculated and VE derived (Noah 1976).
The effectiveness of vaccination, but not its efficacy, can sometimes be checked crudely by observing the correlation between the incidence of disease in separate districts with the vaccination rates in those districts (Sutherland and Fayers 1971; Pollard 1980; Noah and Fowle 1988). Sometimes chance plays a part in affording an opportunity to estimate VE, as in 1970–71 in Texarkana, a city straddling the Texas–Arkansas state line. The Texas side differed from the Arkansas side only in not having a measles vaccine policy, so that during the outbreak of measles, the attack rate (AR) in the Texas side of the city was 48.2 per cent compared with 4.2 per cent in the Arkansas side. VE, based on ARs of 105.9 per thousand and 4.3 per thousand in the unimmunized and in the immunized respectively, was 95.9 per cent (Landrigan 1972).
Case–control studies in vaccination assessments are not common. The design of a case–control vaccine study is similar to that of any other case–control study. Each case of the disease is compared with one or more matched controls without the disease for a history of vaccination against the disease. As in all case–control studies, ‘first cases must be representative of all cases in a specified population with respect to the exposure of interest, and controls must be similarly representative of all non-cases in the same specified population. Second information about exposures and other characteristics must be similar in quality for both cases and controls’ (Comstock 1994). These specifications are extremely difficult to achieve in vaccine case–control studies. The advantages of a case–control study on the other hand are that it can be done more cheaply, quickly, and with fewer cases than for a cohort study. Moreover, the efficacy of vaccine given many years before can be assessed (Smith 1988) and this approach was used recently in a study in the Gambia to assess the efficacy of three or more doses of trivalent oral polio vaccine (Deming 1992). Nevertheless, bias can occur as the vaccines will not have been administered at random (as, for example, if those of higher SE class have higher vaccination rates but lower expectancy of disease), and the vaccine histories in cases and controls may not be accurate (Smith 1988). The problems that need to be given attention in the design of case–control studies are similar to those in cohort vaccine studies. Comstock (1994) has dealt recently with this subject in some detail and Rodrigues and Smith (1999) have written a useful review of the case–control approach to vaccine evaluation.
The case–control method, nevertheless, lends itself well to BCG efficacy studies (Miceli et al. 1988; Smith 1988). First, when the overall incidence of the disease is low, the case–control approach can be used on existing cases of disease to evaluate a vaccine given many years earlier, making it cheaper, quicker, and easier to do than a cohort study. Second, because of the expense of large cohort studies and the time needed to assess the efficacy of BCG, the case–control approach is a more practical one. Third, the difficulty of a ‘double-blind’ approach to a vaccine such as BCG, and fourth, the large variation (from 100 to 57 per cent) in estimates of efficacy of BCG with randomized controlled trials, underline the disadvantages of the cohort approach in trials of BCG.
Calculation of vaccine efficacy
VE is the ratio of the observed diminution in ARs, that is, the difference in ARs between vaccinated and unvaccinated groups, to the expected AR, in the unvaccinated group. Thus
VE = (l – RR) × 100 per cent
where RR is the ratio of AR(vaccinated) to AR(unvaccinated), or relative risk.
In a case–control study, using unmatched controls and odds ratios (OR), results can be arranged as shown in Table 1. Then
VE = (1 – OR) × 100 per cent
where OR = ad/bc.
Table 1 Arrangement of results in case–control study
Refinements of these basic formulae can be found in Orenstein et al. (1985) and the impact of various types of bias on VE results is discussed by Orenstein et al. (1988).
Monitoring of side-effects of vaccines
Monitoring side-effects of vaccines begins at the very first phase of vaccine trials, running in parallel with estimations of efficacy. With randomized controlled trials, the evaluation of side-effects after vaccination is fairly straightforward, it being usually sufficient to compare the incidence of any reactions in cases with that in controls. Where it is not practical, possible, or ethical to conduct a placebo-controlled study, some care is necessary in the interpretation of information on side-effects. In uncontrolled cohort studies, local side-effects (such as a stiff or painful arm) can still be evaluated successfully, but other symptoms, such as fever or convulsions that are non-specific to vaccines and may commonly occur from other causes in healthy children, are more difficult to evaluate without a control group. The time between the vaccination and the appearance of a side-effect also needs to be evaluated carefully. In the earliest trials of MMR vaccine meningitis caused by the mumps component was not documented until the subjects were followed up for at least 28 days (Miller et al. 1993).
In the early measles vaccine trials in England (Measles Vaccine Committee of the Medical Research Council 1966) 9577 children aged 10 months to 2 years were immunized. Eighteen children developed convulsions, 11 of them between post-vaccination days 6 and 9. Only five convulsions were reported in the control group of 16 327 children, none of them between days 6 and 9. This study established not only that convulsions were a real side-effect of measles vaccine, but also that they characteristically occurred during a fixed interval after the vaccine. The need to compare this with the incidence of convulsions after natural disease was also apparent, and they were shown to be at least 10 times more common after the disease than after the vaccine (Miller 1978). This was an extremely important finding in support of the vaccine and illustrates the value of careful epidemiological work in the study of vaccines. In a more recent example (Miller et al. 1993), a side-effect of mumps vaccine (aseptic meningitis) was discovered to be more common than expected (1 in 11 000 doses compared with 4 per million) and to be associated with a vaccine derived from a particular strain (Urabe). This was only discovered after the vaccine was licensed. It was subsequently and promptly withdrawn.
Unlike drug trials, the side-effects of a vaccine can sometimes be evaluated not by using a placebo in the control group but by giving a combination of vaccines omitting the vaccine under trial. The side-effects of pertussis vaccine have been investigated by comparing symptoms in the case group after diphtheria–tetanus–pertussis vaccine (DTP) with those after DT vaccine in the controls (Pollock et al. 1984). Even in this controlled study it was found that adverse publicity to pertussis vaccine led to a reporting bias in the cases given DTP.
Various ingenious studies have been set up to investigate the incidence of rare side-effects after vaccination. In Denmark, Melchior (1977) compared the age-specific incidence of infantile spasms during the period when the whooping cough vaccine was given at 5 months, 6 months, and 15 months with that during the period when it was given at 5 weeks, 9 weeks, and 10 months, and found no difference. In England the extremely rare alleged side-effects of serious and permanent (but non-specific) brain damage after whooping cough vaccine (with wildly varying quoted rates of 1 in 50 000 to 1 in a million) were difficult to refute, mainly because of its rarity, and the lack of a recognizable and confirmable syndrome. The publicity engendered by this led to a dramatic fall in the pertussis vaccine uptake rate from about 79 per cent in the early 1970s to 31 per cent in 1975. A carefully controlled case–control study, the National Childhood Encephalopathy Study, was conducted. In this study the presence or absence of a history of recent vaccination was obtained in all cases of ‘encephalopathy’ reported to the study team and compared with controls. An association was shown between DTP given 3 to 7 days earlier and encephalopathy, but not between DT given at the same time and encephalopathy. It was possible to estimate from this study that the risk of persistent neurological damage in previously healthy children after pertussis vaccine was 1 in 310 000 immunizations, but the 95 per cent confidence limits were very wide, from 1 in 54 000 to 1 in 5.31 million. The reader is referred to the original study (Department of Health and Social Security 1981) for further details of methodology, and to Miller et al. (1982) for a balanced review of the whooping cough vaccine controversy.
Other forms of side-effect evaluation include postmarketing surveillance, which may be active or passive (Stetler et al. 1987). An active surveillance system is less prone to bias, but is likely to be too expensive and unwieldy; the passive system is more prone to bias, but clearly simpler and cheaper. The record linkage method of postlicensure safety assessment also promises to be a useful and reliable method for the routine surveillance of vaccine safety (Miller et al. 1998), as rare side-effects sometimes only become apparent late, often after large well-conducted controlled trials have been successfully completed.
Epidemiological side-effects of vaccination programmes
Mass immunization is mass interference with an ecological process, and inevitably there are side-effects. The most common of these are changes in the age distribution of the infection. The natural cycles of infection are often altered. Usually, if the mass immunization programme involves infants and children, all that this means is a shift of the age distribution to an older age group, without a sustained accompanying increase in incidence in that age group. Outbreaks of rubella and measles in young adults have been reported. This is because if less than ideal coverage occurs or a poor vaccine is used for a mass programme, herd immunity will ensure that fewer people who are susceptible will be exposed to the virus. A group of susceptible persons will then form a ‘cohort’ who may contract the disease in later life.
In Greece, congenital rubella increased in incidence (Panagiotopoulos et al. 1999)—indeed it was reported as an ‘epidemic’ which occurred in 1993. This was attributed to a sustained poor coverage of less than 50 per cent with MMR vaccine in 1-year-old children. The reported incidence of congenital rubella was 24.6 per 100 000 births in 1993. This was the reported incidence in a single year and the authors did not estimate how much lower than expected was the incidence in the intervening years. Even in this situation it is unlikely that a truly sustained increase in incidence occurred over a long period.
Uptake and implementation of vaccines
One has to be a stranger in Jerusalem not to realise that public acceptance of an immunisation procedure determines its success or failure. (Cohen 1978).
The need for field studies in vaccines does not cease at efficacy studies. The implementation of the vaccine is as important as its production, and an effective safe vaccine with poor uptake is of hardly greater benefit than no vaccine at all. In recent years this has been recognized and field studies to investigate the factors that influence uptake of vaccines are now common.
Studies on uptake may measure factors associated with social or service conditions. In one study (Jarman et al. 1988), social conditions associated with being underprivileged (such as overcrowding, unmarried, or single parents), living in high population density areas, being unskilled, and belonging to certain ethnic groups were found to be linked to low immunization uptake. Reasons given by family (Clarke 1980; Blair et al. 1985; Lakhani et al. 1987) or by clinic staff (Adjaye 1981; Lakhani et al. 1987) for refusal can also be investigated. Single-handed GPs, those aged over 65 years, or those with large list sizes and less than average expenditure on community health services were also associated with low uptake rates (Jarman et al. 1988). The efficiency of organization, quality of premises, and adequacy of staffing of three different types of immunization clinic—GP, health centre, and child health clinic—were found to affect compliance (Alberman et al.1986).
In another study the provision of individual performance indicators was enough to stimulate GPs to improve rates (Newlands and Davies 1988). Smith et al. (1976) noted an interesting phenomenon with an annual influenza vaccination programme in industry: in successive years the uptake of vaccine fell considerably. The reasons for this were fairly complex and were probably associated with the need for giving vaccine yearly, the low incidence of influenza during the periods over which the vaccine was given, and the general perception by the target population that the vaccine was not very effective. The message that emanates from many of these studies, however, is that it is the administrative efficiency of the immunization programme and the motivation of the professionals that are the critical features in improving uptake (Noah 1982; Lakhani et al. 1987). Communication strategies can be designed to persuade people to attend (Hingson 1974) and are best used for immunization campaigns.
Evaluation of factors affecting vaccine programmes
The evaluation of the outcome of a vaccine programme can be conducted in several ways. Except with the most successful programmes reasons for non-acceptance of vaccine can be investigated. Reasons for failure/poor uptake may be because of social and professional attitudes to the vaccine, including the effect of the media, or the administrative competence/managerial abilities of those responsible for the programme.
Social and professional attitudes to a vaccine or a vaccine programme may ‘make or break’ the programme. Tenuous evidence that a vaccine may cause serious side-effects, when given the appropriate publicity, can be enough to reduce uptake considerably. An appropriate label for this phenomenon, for it has occurred more than once, is the ‘cry-wolf syndrome’. In the United Kingdom, reports by Kulenkampff et al. (1974) of alleged brain damage following the whooping cough vaccine, fuelled by a television programme in the same year, and by ‘the writings in the lay press and medical journals and media interviews of a professor who was a ‘leading critic of the vaccination policy’ (Cherry 1986) led to a catastrophic fall in pertussis vaccine uptake rates from about 79 per cent between 1967 and 1974 to a low of 30 per cent in 1978. Television programmes were found to be especially influential (McKinnon 1979). The controversy had an effect on other routine vaccinations (DT and poliomyelitis), which fortunately was transient and relatively slight. A survey of professional attitudes during the episode (Wilkinson et al. 1979) suggested that GPs, clinic doctors, or health visitors were ambivalent towards the pertussis vaccine. Most had noticed an increase in parental concern about the vaccine, which was attributed to ‘irresponsible’, ‘ill-controlled’, or ‘biased’ publicity. Moreover, this phenomenon occurred in the United Kingdom without the often-cited problem of litigation risk in the United States.
Adjaye (1981), however, found that parental attitudes towards immunization were greatly influenced by medical and non-medical members of the health professions, while Berkeley (1983) in a study of attitudes to measles vaccine found that the professionals themselves were unsure about the contraindications to vaccine and to its value. In another study (Guest et al. 1986) many professionals offered invalid reasons for failure to give immunization.
Campbell (1983), in analysing reasons for poor uptake of the measles vaccine in Britain, suggested that attitudes and ignorance, a cumbersome policy-making bureaucracy, the absence of legislation (which helped the United States programme) and of a standard record card for each child, and the initiative and motivation required by mothers towards a vaccine given some months after the primary course, were factors that accounted for the unpopularity of measles vaccine in Britain. Encouraging the health professionals to take greater initiative was found to benefit uptake (Carter and Jones 1985). These and other studies (Bussey and Harris 1979; Pugh and Hawker 1986; Lakhani et al. 1987) suggested that education of health professionals and an efficient administration with computerized recall and records were important factors in improving uptake rates. In more recent years in Britain, a more efficient system for the administration of vaccination programmes, together with the rapid evaluation of immunization rates (the COVER programme) (Begg et al. 1989) have helped considerably in raising MMR vaccine immunization rates to well over 90 per cent. With the computerized COVER programme, districts with poor immunization rates can be targeted. The addition of the mumps and rubella components to the existing measles vaccine has undoubtedly helped also. Indeed this emphasis on efficient administration has since brought the vaccine uptake rates to acceptable levels. In the United States, school immunization laws were also a proven method of improving immunization uptake (Robbins et al. 1981), although its value should not be taken for granted in other countries.
Unfortunately, the ‘cry-wolf’ syndrome has reappeared in Britain to affect measles vaccine uptake. This arose from some weak evidence that Crohn’s disease and childhood autism may be associated with measles vaccine (Thompson et al. 1995), even though there is good evidence that either association is unlikely (Miller and Waight 1998; Pebody et al. 1998; Taylor et al. 1999, 2000).
Evaluation of outcome of immunization programmes
The case for long-term serological surveillance of immunization programmes has been forcefully advocated by Evans (1980) and Raska (1971). Evans argued that a surveillance system based on the reporting of disease alone was insufficiently sensitive or specific to monitor a vaccine programme. An individual’s history of immunization or disease correlated well with antibody presence in measles and mumps, but less well in rubella and poliomyelitis; moreover, for tetanus natural infection has virtually no bearing on antibody levels.
The reliability of a negative history of disease or immunization was, however, poor for measles, mumps, and rubella. The serological demonstration of satisfactory levels of durable antibody in immunized persons was a more reliable measure of vaccine effectiveness than monitoring the incidence of the disease itself, and absence of antibody in particular communities or specific age groups could identify those who may need protection. Long-term surveillance of immunity after measles vaccination has also been advocated (Anonymous 1976). Serological studies have shown that vaccine failure and not waning vaccine-induced antibody accounted for most of the low or undetectable antibody titres in immunized children, especially those immunized before 13 months of age (Yeager et al. 1977).
A study of measles vaccine on the survival pattern of 7- to 35-month-old children in Kasongo, Zaire (Kasongo Project Team 1981) used life-table analysis to evaluate outcome of the vaccination programme in a community with a high measles incidence and measles case-fatality rate. They found that, although the measles vaccine programme undoubtedly reduced the risk of measles death, overall survival was less influenced by measles vaccine after 22 months of age.
Another study found that the success of a vaccine programme against measles, mumps, and rubella using the MMR vaccine was reflected in the virtual disappearance of a common complication of all three infections (encephalitis) in their population of children (Koskiniemi and Vaheri 1989). Improved intellectual performance was an outcome noted in a cohort study in children who had been immunized against pertussis compared with those who had been hospitalized for the disease, even after allowing for social differences in the two groups (Butler et al. 1982).
Costing studies of vaccines
Costing studies are important both in assessing the need for a vaccine and in evaluating its efficiency. The detailed methodology of costing studies for infectious disease is considered elsewhere in this book. Costs include costs of the vaccine and its administration. Costs of the vaccine may be influenced by factors as diverse as costs of development and market competition: the price of human-derived hepatitis B vaccine halved when the yeast-derived vaccine began to be marketed. Administration costs will include the cost of the syringe, personnel, accommodation as well as costs of initiating and administering a record and follow-up programme. Benefits include savings on treatment costs, mortality, morbidity, the avoidance of intangibles such as pain and grief, and external benefits (Creese and Henderson 1980). It is debatable whether social benefits of successful intervention should be costed, because the value of such benefits may not be convincing to pragmatic health-care managers. Patrick and Woolley (1981) have pointed out that health managers may fail to be impressed by cost–benefit studies because the costs, which are health costs, have to be weighed against benefits that are usually social. Nevertheless, the benefits of providing supportive cost calculations are an important addition to the valuation of a vaccination programme.
A list of the benefits of immunization, simply and clearly stated, often cannot fail to convince as in one of the earliest such papers (Witte and Axnick 1975) describing the results of 10 years of measles immunization in the United States in terms of lower morbidity and mortality, and improved quality of life (Table 2).
Table 2 Benefits of measles immunization over 10 years (United States)
Other studies (Creese and Henderson 1980; Koplan 1985) have amply shown the value for money of measles vaccine in terms of benefit–cost ratios for measles vaccine ranging from 10:1 to 15:1. Cost studies can be successfully performed in general practice (Binnie 1984). In this general practice, over a 20-year period, immunization not only brought about a small financial reward, but also reduced the number of consultations by 40 per cent, even though a home visit was often necessary to immunize a child. Examples of cost–benefit analysis of various immunizations that have demonstrated the benefit of vaccines, and to which the reader is referred for details of methodology, include measles vaccine (Witte and Axnick 1975; Albritton 1978), pertussis vaccine (Koplan et al. 1979; Hinman and Koplan 1984, 1985), mumps vaccine (White et al. 1985; Koplan and Preblud 1982), poliomyelitis (Fudenberg 1973), pneumococcal vaccine (Patrick and Woolley 1981), and hepatitis B in homosexuals (Adler et al. 1983). General articles on cost–benefits of immunization programmes include Creese and Henderson (1980) and Koplan (1985). A formula has been developed for calculating vaccine profitability, defined as the economic yield, positive or negative, obtained per monetary unit invested in a campaign (Carvasco and Lardinois 1987).
This section shows that the methods used in the study of the epidemiology of infection are similar to those used for any other type of disease, but are also sufficiently different to warrant special consideration. This is best illustrated in the section on vaccines. Surveillance, first used in infectious disease, has become an essential technique in public health practice generally. The importance of collaboration between epidemiologists and other scientists, discussed here in relation to the microbiologist, is relevant to any epidemiological specialism.
The epidemiologist and the microbiology laboratory
Close collaboration between infectious disease epidemiologists and microbiologists is essential for effective outbreak investigation and control, for surveillance, and for vaccine use. Moreover, the microbiologist needs the epidemiologist and to have an understanding of epidemiology, as much as the epidemiologist needs the microbiologist and knowledge of microbiology. Microbiology and epidemiology each has strengths and weaknesses that need to be acknowledged; a partnership between the two disciplines is essential in infectious disease epidemiology so that each can complement the other.
Outbreaks often cannot be effectively investigated without the use of epidemiology—in some food poisoning incidents, for example, microbiological technology is still too limited to be able to identify the causal organism in the implicated food or in affected patients. In others, the implicated foods may be unavailable for microbiological testing. Similarly, some outbreak investigations will be unsuccessful without the use of microbiology; in an outbreak of fever and rash in a school in England (McEvoy et al. 1987), the epidemiological investigation was considerably hampered by the lack of a microbiological diagnosis. This outbreak is described in more detail below.
For surveillance of infectious disease, notifications based on clinical diagnosis alone, although useful if laboratory support is not available, are often lacking in precision and detail. In epidemiological studies and monitoring of vaccines, microbiological support is also essential for added precision, including for example, evaluating the effect of vaccine on carriage of an organism. Facets of this collaboration between the epidemiologist and the microbiologist (and indeed with other disciplines, such as entomologists and zoologists) are explored in more detail below.
Microbiology for epidemiologists
Infectious disease epidemiologists require sound technical understanding of infection and micro-organisms to perform their work effectively. Knowledge of the ease or difficulty and expense of undertaking laboratory tests is clearly important. An understanding of what the tests mean, and of their sensitivity and specificity is needed for informed interpretation of the results. The epidemiological significance of typing of organisms also has to be understood, as does the meaning of complex antibody tests, e.g. for hepatitis B virus. A working understanding of immunology is also necessary. Some clinical knowledge of infectious disease is essential for the effective recognition and definition of cases, and for their management.
Epidemiology for microbiologists
Microbiologists likewise should have an understanding of the strengths and limitations of epidemiology to be able to work in partnership with epidemiologists. This applies equally both to those microbiologists working in the community and in the hospital. Massive uncoordinated sampling of patients and/or food and the environment can usually be avoided in outbreak investigation. Knowledge of sensitivity and specificity, especially applied to laboratory testing, is essential. For surveillance the importance of routine reporting following clearly defined rules must be recognized.
Aspects of the collaboration between epidemiologist and microbiologist are considered under the subjects of surveillance, outbreaks, and vaccines.
The laboratory in surveillance
Collaboration between microbiologist and epidemiologist undoubtedly leads to more sensitive and more refined surveillance, widening its scope considerably. Notifications of infectious disease, although essential, at best are usually fairly crude indicators of infection.
Confirmation of diagnosis
Clinical diagnoses, normally the basis for notifications, are not always accurate. Clinical malaria in south-eastern United States was shown by laboratory investigation not to be malaria when the Centers for Disease Control was first established in 1941; this early example underlined the importance of the laboratory in surveillance (Langmuir 1963). Rubella is notoriously difficult to diagnose clinically; it is often confused with Fifth disease, for example, and a laboratory surveillance system that includes both rubella virus and parvovirus B19 will distinguish between these. In a successful vaccine programme, as the disease reaches low levels, it is often necessary to confirm each notified case by using the laboratory. Otherwise it is impossible to judge the success of the intervention with any confidence. In the WHO programme for the world eradication of poliomyelitis, the laboratory is essential for confirming the diagnosis even in this easily recognizable and characteristic disease. For WHO to ratify eradication (Wright et al. 1991), a country has to continue surveillance with laboratory backup for 3 years before the last confirmed case.
Reaching precision in diagnosis
Most notifiable diseases lack precision, which the laboratory can give. Influenza covers several types, subtypes and variants (Table 3), which are generally indistinguishable clinically. Food poisoning covers not only several different types of infection, but often each of these infections can also be typed or subtyped further. Thus the causes of food poisoning include salmonellas, Clostridium perfringens, staphylococci, and Escherichia coli. All those organisms can be typed, and the more common salmonellas can be further phage-typed to afford even greater precision in diagnosis. The use of molecular techniques in diagnosis has expanded considerably in recent years and can be used for surveillance as well as outbreak investigation. Antibiotic resistance patterns can also be used for surveillance. Table 3 gives some examples of the precision and detail that surveillance underpinned by the laboratory can provide. Some of this detail is only useful for precision within outbreaks, e.g. phage-typing of staphylococci and C. perfringens; others are useful for surveillance—salmonellas, influenza, and meningococcal meningitis.
Table 3 Examples of detail provided by the laboratory
Increasing the scope of surveillance
There are many infections for which only laboratory surveillance is feasible—these include legionnaires’ disease, psittacosis and Q fever, most adenoviral infections, and the enteroviruses (coxsackie A and B, echoviruses). Many of these infections are important public health problems, especially in better-developed countries in which the more traditional epidemic infections found in developing countries have become less common. Most of these infections are important enough to require surveillance to underpin outbreak investigation. Psittacosis and Q fever are zoonoses, and the enteroviruses cause outbreaks of lymphocytic meningitis, Bornholm disease, and other clinical conditions in the autumn season. The environmental sources of legionnaires’ disease—mainly water—are well known.
Laboratory diagnosis is important for international surveillance. The spread of Vibrio cholerae el Tor across the world has been well-documented (Anonymous 1993). International surveillance of the subtypes and variants of influenza A virus is critical both for documenting spread and for early warning, and for the formulation of vaccines (Gensheimer et al. 1999; Snacken et al. 1997). Further examples of important infections in which laboratory backup of surveillance plays a key part in control are dengue, HIV-1 and HIV-2, and malaria.
Vectors and reservoirs of infection
The use of the laboratory in tracking the origin and spread of infection in animals and arthropods, as well as in water and soil, is important but will not be discussed here in detail. Influenza virus is of particular interest because of the interchanges and evolution of infection between a variety of animals and humans. The possible origin of new influenza subtypes in the pig and duck populations and its relationship to their agricultural environment of parts of China are still being studied (Webster et al. 1992), and the concern caused by avian influenza virus crossing the species barrier from chickens to humans (Snacken et al. 1997) was worldwide. The existence of a ‘zoonotic pool’ for viruses (Morse 1993) clearly justifies close epidemiological and microbiological surveillance to continue for the foreseeable future.
The laboratory in epidemiological investigation of vaccines
Technical microbiological expertise is clearly crucial to the development and preparation of vaccines, and will not be considered further here. The use of the microbiological laboratory will be considered more fully in the section on the field investigation of vaccines in this chapter, so will only be considered in outline.
In phase 1 and 2 trials, efficacy is usually assessed by measurement of antibody, which requires serological testing. In phase II trials in particular, phasing of vaccine doses will depend on antibody levels found. In field trials (phase 3) microbiology is usually required to confirm diagnosis of cases occurring in both case and control groups, and this may be done by isolation and typing of the organism, or by serology, or both.
As a vaccine programme develops the overall incidence will fall, but cases will continue to occur for some time and microbiological confirmation of diagnosis becomes more important as the incidence diminishes to very low levels. In the present WHO initiative to eradicate poliomyelitis from the world, laboratory facilities worldwide needed strengthening to investigate every reported case of acute flaccid paralysis (Wright et al. 1991). Only when every reported case of acute flaccid paralysis proves not to be poliomyelitis over a period of 3 years can there be confidence that eradication has been achieved.
Assessment of vaccine success by serological testing of populations will be described in the section on the field investigations of vaccines.
The laboratory in outbreak investigation
In no other situation is the partnership between microbiologist and epidemiologist more important perhaps than in the investigation of an outbreak. Some outbreaks—such as food-borne outbreaks of small round structured virus infection or hepatitis A—can really only be solved by epidemiological techniques, as the causative organisms cannot easily be detected in food. In others, especially those food-borne outbreaks in which the menu does not offer a choice, so that everyone usually eats everything on offer, the epidemiologist is considerably restricted, and a microbiological solution is often the only hope.
Inevitably, in many outbreaks no foods are available for testing, so that epidemiological analysis is essential. In other outbreaks too many foods are available for microbiological analysis: in a continuing outbreak of gastroenteritis on board ship, investigated by the PHLS Communicable Disease Surveillance Centre, London (O’Mahony et al. 1986), no particular meal was obviously at fault. More than 400 food items and ingredients were available for testing and a microbiological approach would have meant that all these had to be tested, without any guarantee of success Careful epidemiological analysis of the outbreak by time, place, and person showed that drinking water was the probable cause of the outbreak, and subsequent investigation of the water tanks revealed a defect through which the water was probably contaminated by sewage outfall from the ship. In other food poisoning outbreaks, the epidemiological proof can be overwhelming, although the level of contamination in the offending food is so low that it is only confidence in the epidemiology that enables the laboratory to persist in ensuring ‘microbiological proof’. In an outbreak of Salmonella ealing infection caused by powdered infant milk, laboratories sampled more than 4000 unopened milk cartons before the organism was isolated (Rowe et al. 1987).
The outbreak of Streptobacillus moniliformis infection in a girls’ school in rural England illustrated well the need for microbiological and epidemiological expertise (McEvoy et al. 1987). Until the microbiological diagnosis was made, the epidemiologists were attempting to analyse a large number of risk factors found in a rural and school environment. When the illness was diagnosed, a review of the literature revealed that one previous outbreak of the infection had been described in Haverhill, Massachusetts (hence the alternative name of Haverhill fever for the illness), and had been attributed to raw milk. The girls in the school had indeed been exposed to raw milk. However, subsequent careful epidemiological analysis showed water to be the cause. An examination of the water supply revealed that water from a pond in the school grounds, which was infested with rodents, was contaminating the school drinking water supply after it had been chlorinated.
Microbiology is useful for confirming foods and other sources of infection in outbreaks, such as water tanks (Campylobacter) and cooling towers (Legionella pneumophila), birds (psittacosis), and animals (rabies, plague). Increasing precision in microbiological diagnosis—typing of organisms such as phage typing and more recently molecular techniques, as well as typing of toxins—has clearly been useful. In an outbreak of AIDS attributed to a dentist in Florida, molecular typing showed identical patterns in the virus obtained from patients and the primary case, thus confirming beyond reasonable doubt that he had indeed caused the infection in his patients (Hillis and Huelsenbeck 1994).
The epidemiologist can on the other hand aid the microbiologist in giving guidance on the number of patients to test in an outbreak, and the foods or other items that need to be sampled. The pattern of the outbreak is also crucial to its management—in propagated or case-to-case outbreaks, e.g. a salmonella outbreak in a hospital, a planned epidemiological and microbiological approach to both investigation and management is the most efficient option. Point source outbreaks on the other hand tend to create more publicity but are on the whole easier to control. The epidemiologist’s role in differentiating between these two types of outbreaks is essential.
The work of the microbiologist in assessing levels of contamination in food is important to management and prevention. Levels of salmonella contamination in eggs have been shown to be extremely low, and the organisms have been discovered in both the white and the yolk, either together or separately (Humphrey et al. 1989). This type of information ensures that correct advice on prevention can be given for handling and cooking eggs.
In a successful partnership there is always a need for standard operating procedures, which should be drawn up carefully and followed by all participants. These should include protocols for surveillance, covering the reporting of results (when, what, and how) by the laboratory and feedback by the epidemiologists. Standards for laboratory reporting and for epidemiological analysis, and the development and roll-out of new tests as they come into routine use, should be covered. Confidentiality issues on both sides should also be carefully laid down and the appropriate safeguards spelled out.
Adjaye, N. (1981). Measles immunization. Some factors affecting non-acceptance of vaccine. Public Health, London, 95, 185.
Adler, M.W., Belsey, E.M., McCutchan, J.A., et al. (1983). Should homosexuals be vaccinated against hepatitis B? Cost and benefit assessment. cBritish Medical Journal, 286, 1621.
Alberman, E., Watson, E., Mitchell, P., et al. (1986). The development of performance and cost indicators for preschool immunization. Archives of Disease in Childhood, 61, 251.
Albritton, R.B. (1978). Cost-benefits of measles eradication: effects of a federal intervention. Policy Analysis, 4, 1.
Anonymous (1976). Leading article: vaccination against measles. Lancet, ii, 132–4.
Anonymous (1993). Cholera—update end of 1993. Weekly Epidemiological Record, 69, 13–20.
Arvin, A.M. (1997). The varicella vaccine. Current Clinical Topics in Infectious Diseases, 17, 110–46.
Begg, N.T., Gill, O.N., and White, J. (1989). COVER (cover of vaccination evaluation rapidly): description of the England and Wales scheme. Public Health, 103, 81.
Berkeley, M.I.K. (1983). Measles—the effect of attitudes on immunisation, 41/3, pp. 141–7. Health Bulletin. Scottish Home and Health Department, Edinburgh.
Binnie, G.A.C. (1984). Measles immunization profit and loss in a general practice. British Medical Journal, 289, 1275.
Blair, S., Shave, N., and McKay, J. (1985). Measles matters, but do parents know? British Medical Journal, 290, 623.
Boussard, E., Flahault, A., Vibert, J-F., and Valleron, A.J. (1996). Sentiweb: French communicable disease surveillance on the world-wide web. British Medical Journal, 313, 1381–4.
Bradford Hill, A. (1971). Principles of medical statistics (9th edn), Chapter 24. Lancet, London.
Bussey, A.L. and Harris, A.S. (1979). Computers and effectiveness of the measles vaccination campaign in England and Wales. Community Medicine, 1, 29.
Butler, N.R., Golding, J., Haslum, M., et al. (1982). Recent findings from the 1970 child health and education study: preliminary communication. Journal of the Royal Society of Medicine, 75, 781.
Campbell, A.G.M. (1983). Measles immunization: why have we failed? Archives of Disease in Childhood, 58, 3.
Came, S. (1984). Place of development surveillance in general practice. Journal of the Royal Society of Medicine, 77, 819.
Carter, H. and Jones, I.O. (1985). Measles immunization: results of a local programme to increase vaccine uptake. British Medical Journal, 290, 1717–19.
Carvasco, J.L. and Lardinois, R. (1987). Formula for calculating vaccine profitability. Vaccine, 5, 123–7.
Cherry, J.D. (1986). The controversy about pertussis vaccine. Current Clinical Topics in Infectious Diseases, 7, 216.
Clarke, S.J. (1980). Whooping cough vaccination: some reasons for non-complction. Journal of Advanced Nursing, 5, 313–19.
Cochi, S.L., Preblud, S.R., and Orenstein, W.A. (1988). Perspectives on the resurgence of mumps in the United States. American Journal of Diseases of Children, 142, 499–507.
Cockburn, W.C. (1969). World aspects of the epidemiology of rubella. American Journal of Diseases of Children, 118, 112.
Cohen, H. (1978). Vaccination against pertussis, yes or no? In International symposium on pertussis (ed. C.R. Manclark and J.C. Hill), p. 249. National Institutes of Health, Bethesda, MD.
Collete, B.J.A. (1982). The sentinel practices system in The Netherlands. In Environmental epidemiology (ed. P.E. Leaverton), p. 149. Praeger, New York.
Comstock, G.W. (1994) Evaluating vaccination effectiveness and vaccine efficacy by means of case-control studies. Epidemiologic Reviews, 16.
Cowden, J.M., O’Mahoney, M., Bartlett, C.L.R., et al. (1989). A national outbreak of Salmonella typhimurium DT 124 caused by contaminated salami sticks. Epidemiology and Infection, 103, 219.
Creese, A.L. and Henderson, R.H. (1980). Cost-benefit analysis and immunization programmes in developing countries. Bulletin of the World Health Organization, 58, 491.
Cutting, W.A.M. (1983). Measles immunization. A review. Journal of Tropical Pediatrics, 29, 246–7.
Deming, M., Jaiteh, K.O., Otten, M.W., et al. (1992) Epidemic poliomyelitis in the Gambia following the control of poliomyelitis as an endemic disease. 11. Clinical efficacy of trivalent oral polio vaccine. American Journal of Epidemiology, 135, 393–408.
Department of Health and Social Security (1981). Whooping cough. Reports from Committee on Safety of Medicines and the Joint Committee on Vaccination and Immunization. HMSO, London.
Epidemiology Research Laboratory (1975). Echovirus 19 this summer? British Medical Journal, ii, 346.
Evans, A.S. (1980). The need for serologic evaluation of immunization programs. American Journal of Epidemiology, 112, 725.
Eylenbosch, W.J. and Noah, N.D. (ed.) (1988). Surveillance in health and disease. Oxford University Press.
Fine, P.E.M., Clarkson, J.A., and Miller, E. (1988). The efficacy of pertussis vaccines under conditions of household exposure. International Journal of Epidemiology, 17, 635.
Fleming, D.M. and Crombie, D.L. (1985). The incidence of common infectious diseases: the weekly returns service of the Royal College of General Practitioners. Health Trends, 17, 13.
Fox, J.L. (1996). ….but new plans for surveillance are still being proposed. Nature Medicine, 2, 733.
Fudenberg, H.H. (1973). Fiscal returns of biomedical research. Journal of Investigative Dermatology, 61, 321.
Gensheimer, K.F., Fukuda, K., Brammer, L., et al. (1999). Preparing for pandemic influenza: the need for enhanced surveillance. Emerging Infectious Diseases, 5, 297–9.
Gill, O.N., Bartlett, C.L.R., Sockett, P.N., et al. (1983). Outbreak of Salmonella napoli infection caused by contaminated chocolate bars. Lancet, i, 574.
Gleason, R.E., Khan, C.B. Funk, I.B., and Craighead JE. (1982). Seasonal incidence of insulin dependent diabetes (IDDM) in Massachusetts 1964–1973. International Journal of Epidemiology, 11, 39–45.
Guest, M., Horn, J., and Archer, L.N.J. (1986). Why some parents refuse pertussis immunisation. Practitioner, 230, 210.
Hall, S.M. and Glickman, M. (1988). The British paediatric surveillance unit. Archives of Disease in Childhood, 63, 344–346.
Haskey, J.C. and Birch, D. (1985). Statistics from general practice: morbidity and its measurement using practice statistical reports. Health Trends, 17, 32.
Helweg-Larsen, K., Bay, H., and Mac, F. (1985). A statistical analysis of the seasonality in sudden infant death syndrome. International Journal of Epidemiology, 14, 566–574.
Henderson, D.A. (1976). Surveillance of smallpox. International Journal of Epidemiology, 5, 19–28
Henderson, R.H. and Sundaresan, T. (1982). Cluster sampling to assess immunization coverage: a review of experience with a simplified sampling method. Bulletin of World Health Organization, 60, 253–60.
Hillis, D.M. and Huelsenbeck, J.P. (1994) Support for dental HIV transmission. Nature, 369, 24–5.
Hingson, R. (1974). Obtaining optimal attendance at mass immunization programs. Health Services Reports, 89, 53–64.
Hinman, A.R. and Koplan, J.P. (1984). Pertussis and pertussis vaccine. Reanalysis of benefits, risk and costs. Journal of the American Medical Association, 251, 3109–13.
Hinman, A.R. and Koplan, J.P. (1985). Pertussis and pertussis vaccine: further analysis of benefits, risks and costs. Development in Biological Standardization, 61, 429–37.
Hinman, A.R., Brandling-Bennett, A.D., Bernier, R., et al. (1980). Current features of measles in the United States: feasibility of measles elimination. Epidemiologic Reviews, 2, 153–70.
Humphrey, T.J., Baskerville, A., Mawer, S., Rowe, B., and Hopper, S. (1989). Salmonella enteritidis phage type 4 from the contents of intact eggs: a study involving naturally infected hens. Epidemiology and Infection, 103, 415–23.
Irwig, L.M. (1976). Surveillance in developed countries with particular reference to child growth. International Journal of Epidemiology, S57–61.
Jarman, B., Bosanquet, N., Rice, P., et al. (1988). Uptake of immunization in district health authorities in England. British Medical Journal, 296, 1775–8.
Judelsohn, R.G., Pleissner, M.L., and O’Mara, D.J. (1980). School-based measles outbreaks: correlation of age at immunization with risk of disease. American Journal of Public Health, 70, 1162–5.
Kasongo Project Team (1981). Influence of measles vaccination on survival pattern of 7–35-month-old children in Kasongo, Zaire. Lancet, i, 764–7.
Khot, A. and Burn, R. (1984). Seasonal variation and time trends of deaths from asthma in England and Wales (1960–1982). British Medical Journal, 289, 233–4.
Klaucke D.N. (1994) Evaluating public health surveillance. In Principles and practice of public health surveillance (ed. S.M. Teutsch and R.E. Churchill). Oxford University Press.
Koplan, J.P. (1985). Benefits, risks and costs of immunization programmes. In The value of preventive medicine, Ciba Foundation Symposium 110, pp. 55–680. Pitman, London.
Koplan, J.P. and Preblud, S.R. (1982). A benefit–cost analysis of mumps vaccine. American Journal of Diseases of Children, 136, 362.
Koplan, J.P., Schoenbaum, S.C., Weinstein, M.C., and Fraser, D.W. (1979). Pertussis vaccine—an analysis of benefits, risks and costs. New England Journal of Medicine, 301, 906–11
Kosklniemi, M. and Vaheri, A. (1989). Effect of measles, mumps, rubella vaccination on pattern of encephalitis in children. Lancet, i, 31–4.
Kulenkampff, M., Schwartzman, J.S., and Wilson, J. (1974). Neurological complications of pertussis inoculation. Archives of Disease in Childhood, 49, 46–9.
Lakhani, A.D.H., Morris, R.W., Morgan, M., Dale, C., and Vaile, M.S. (1987). Measles immunization: feasibility of a 90 per cent target uptake. Archives of Disease in Childhood, 62, 1209.
Landrigan, P.J. (1972). Epidemic measles in a divided city. Journal of the American Medical Association, 221, 567–70.
Langmuir, A.D. (1963). The surveillance of communicable diseases of national importance. New England Journal of Medicine, B>268, 182–92.
Langmuir, A.D. (1976). William Farr: founder of modern concepts of surveillance. International Journal of Epidemiology, 5, 13–18.
Last, J.M. (ed.) (1995). A dictionary of epidemiology (3rd edn). Oxford University Press.
Limentani, A.E., Elliott, L.M., Noah, N.D., and lambourn, J. (1979). Outbreak of hepatitis B from tattooing. Lancet, ii, 86–8.
Lynn, R., Nicoll, A., Rahi, J., and Verity, C. (ed.) (1999). British Paediatric Surveillance Unit annual report 1998/9. British Paediatric Surveillance Unit, London.
McCarthy, K. and Taylor Robinson, C.H. (1963). Immunization against measles. British Journal of Clinical Practice, 17, 650–9.
McKinnon, J.A. (1979). The impact of the media on whooping cough immunization. Health Education Journal, 37, 198.
Maclean, M.H. and Macleod, A. (1984). Seasonal variation in the frequency of anencephalus and spina bifida births in the UK Journal of Epidemiology and Community Health, 38, 99–102.
Marks, J.S., Hayden, G.F., and Orenstein, W.A. (1982). Methodologic issues in the evaluation of vaccine effectiveness. American Journal of Epidemiology, 116, 510–23.
McEvoy, M.B, Noah, N.D., and Pilsworth, R. (1987). Outbreak of fever caused by Streptobacillus imoniliformis. Lancet, ii, 1361–3.
Measles Vaccine Committee of the Medical Research Council (1966). Vaccination against measles. British Medical Journal, i, 441–6.
Melchior, J.C. (1977). Infantile spasms and early immunization against whooping cough: Danish survey from 1970 to 1975. Archives of Disease in Childhood, 52, 134–7.
Miceli, I., de Kantor, I., Colalacovo, D., et al. (1988). Evaluation of the effectiveness of BCG vaccination using the case-control method in Buenos Aires, Argentina. International Journal of Epidemiology, 17, 629–34.
Miller, C.L. (1978). Severity of notified measles. British Medical Journal, i, 1253.
Miller, D.L. (1964). Frequency of complications of measles, 1963. British Medical Journal, ii, 75–8.
Miller, D.L., Alderslade, R., and Ross, E.M. (1982). Whooping cough and whooping cough vaccine: the risks and benefits debate. Epidemiologic Reviews, 4, 1–24.
Miller, E. and Waight, P. (1998). Measles, measles vaccination, and Crohn’s disease. British Medical Journal, 316, 1745.
Miller, E., Goldacre, M., Pugh, S., et al. (1993) Risk of aseptic meningitis after measles, mumps, and rubella vaccine in UK children. Lancet, 341, 979–82.
Miller, E., Waight, P., Gay, N., et al. (1997). The epidemiology of rubella in England and Wales before and after the 1994 measles and rubella campaign. CDR Review No. 2, R 26–32.
Miller, E., Waight, P., and Farrington, P. (1998). Safety assessment post-licensure. In Preclinical and clinical development of new vaccines. Vol. 95, Developments in biology standardization (ed. S. Plotkin, F. Brown, and F. Horaud), pp. 235–43. Karger, Basel.
Monto, A.S. (1974). The Tecumseh study of respiratory illness. American Journal of Epidemiology, 100, 458–68.
Moro, M.L. and McCormick, A. (1988). Surveillance for communicable disease. In Surveillance in health and disease (ed. W.J. Eylenbosch and N.D. Noah), p. 166. Oxford University Press.
Morgan-Capner, P., Wright, J., Miller, C.L., et al. (1988). Surveillance of antibody to measles, mumps and rubella by age. British Medical Journal, 297, 770–2.
Morley, D. (1975). Nutritional surveillance of young children in developing countries. International Journal of Epidemiology, 5, 51–5.
Morse, S.S. (1993). Examining the origins of emerging viruses. In Emerging viruses (ed. S.S. Morse). Oxford University Press, New York.
Newlands, M. and Davies, L. (1988). The use of performance indicators for immunization rates in General Practice. Public Health, 102, 269–73.
Noah, N.D. (1974). Mycoplasma pneumoniae infection in the United Kingdom, 1967–1973. British Medical Journal, ii, 544–6.
Noah, N.D. (1976). Attack rates of notified whooping cough in immunized and unimmunized children. British Medical Journal, i, 128–9.
Noah, N.D. (1982). Measles eradication policies. British Medical Journal, 284, 99–8.
Noah, N.D. (1983). The strategy of immunization. Community Medicine, 5, 140–7.
Noah, N.D. (1989). Cyclical patterns and predictability in infection. Epidemiology and Infection, 102, 175–90.
Noah, N.D. and Fowle, S.E. (1988). Immunity to rubella in women of childbearing age in the United Kingdom. British Medical Journal, 297, 1301–4.
O’Mahony, M., Noah, N.D., Evans, B., et al. (1986) An outbreak of gastroenteritis on a passenger cruise ship. Journal of Hygiene, 97, 229–36.
Orenstein, W.A., Bernier, R.H., Dondero, T., et al. (1985). Field evaluation of vaccine efficacy. Bulletin of the World Health Organization, 63, 1055.
Orenstein, W.A., Bernier, R.H., and Hinman, A.R. (1988). Assessing vaccine efficacy in the field. Epidemiologic Reviews, 10, 212–41.
Panagiotopoulos, T., Antoniadou, I., and Valassi-Adam, E. (1999) Increase in congenital rubella occurrence after immunisation in Greece: retrospective survey and systematic review. British Medical Journal, 319, 1462–7.
Patrick, K.M. and Woolley, F.R. (1981). A cost-benefit analysis of immunization for pneurnococcal pneumonia. Journal of the American Medical Association, 245, 473–7.
Pebody, R.G., Paunio, M., and Ruutu, P. (1998). Crohn’s disease has not increased in Finland. British Medical Journal, 316, 1745.
PHLS Epidemiological Research Laboratory and 21 Area Health Authorities (1982). Efficacy of pertussis vaccination in England. British Medical Journal, 285, 357–9.
PHLS Whooping Cough Committee and Working Party (1973). Efficacy of whooping cough vaccines used in the United Kingdom before 1968: final report. British Medical Journal, 1, 259–62.
Pollard, R. (1980). Relation between vaccination and notification rates for whooping cough in England and Wales. Lancet, i, 1180–2.
Pollock, T.M. (1966). Trials of prophylactic agents for the control of communicable diseases. Monograph series No. 52. World Health Organization, Geneva.
Pollock, T.M., Miller, E., Mortimer, J.Y., et al. (1984). Symptoms after first immunization with DTP and DT vaccine. Lancet, ii, 146–9.
Pugh, E.J. and Hawker, R. (1986). Measles immunisation: professional knowledge and intention to vaccinate. Community Medicine, 8, 340–7.
Raska, K. (1971). Epidemiological surveillance with particular reference to the use of immunological surveys. Proceedings of the Royal Society of Medicine, 64, 684–8.
Research Unit of the Royal College of General Practitioners (1974). The incidence and complications of mumps. Journal of the Royal College of General Practitioners, 24, 545–51.
Rey, M. (1985). Eradication of measles by widespread vaccination is beneficial and feasible. Semaine des Hôpitaux de Paris, 61, 21–5.
Robbins, K.B., Brandling-Bennett, A.D., and Hinman, A.R. (1981). Low measles incidence: association with enforcement of school immunization laws. American Journal of Public Health, 71, 270–4.
Rodrigues, L.C. and Smith, P.G. (1999) Use of the case-control approach in vaccine evaluation: efficacy and adverse effects. Epidemiologic Reviews, 21, 56–72.
Rowe, B., Hutchinson, D.N., Gilbert, R.J., et al. (1987). Salmonella ealing infections associated with consumption of infant dried milk. Lancet, ii, 900–3.
Sagliocca, L., Amoroso, P., Stroffolini, T., et al. (1999). Efficacy of hepatitis A vaccine in prevention of secondary hepatitis A infection: a randomised trial. Lancet, 353, 1136–9.
Salmon, R.L. and Bartlett, C.L.R. (1995). European surveillance systems. Reviews in Medical Microbiology, 6, 267–76.
Sever, J.L., Hardy, J.B., Nelson, K.B., et al. (1969). Rubella in the collaborative perinatal research study. American Journal of Diseases of Children, 118, 123–32.
Silverstein, A.M. (1981). Pure politics and impure science: the swine flu qffair. Johns Hopkins University Press, Baltimore, MD.
Smith, J.W.G., Fletcher, W.B., and Wherry, P.J. (1976). Vaccination in the control of influenza. Postgraduate Medical Journal, 52, 399–404.
Smith, P.G. (1988). Epidemiological methods to evaluate vaccine efficacy. British Medical Journal, 44, 679.
Snacken, R., Kendal, A.P., Haaheim, L.R., and Wood, J.M. (1997). The next influenza pandemic: lessons from Hong Kong. Emerging Infectious Diseases, 5, 195–203.
Stetler, H.C., Mullen, J.R., Brennan, J.P., Livengood, J.R., Orenstein, W.A., and Hinman, A.R. (1987). Monitoring system for adverse events following immunization. Vaccine, 5, 169–74.
Sutherland, I. and Fayers, P.M. (1971). Effect of measles vaccination on incidence of measles in the community. British Medical Journal, 1, 698–702.
Taylor, B., Miller, E., Farrington, C.P., et al. (1999). Autism and measles, mumps, and rubella vaccine: no epidemiological evidence for a casual association. Lancet, 353, 2026–9.
Taylor, B., Miller, E., and Farrington, P. (2000). Autism and measles, mumps, and rubella vaccine. Lancet, 355, 409.
Thacker, S.B. (1994) Historical development. In Principles and practice of public health surveillance. (ed. S.M. Teutsch and R.E. Churchill). Oxford University Press.
Thacker, S.B. and Berkelman, R.J. (1986). Public health surveillance in the United States. Epidemiologic Reviews, 10, 164–90.
Thacker, S.B., Choi, K., and Brachman, P.S. (1983). The surveillance of infectious diseases. Journal of American Medical Association, 249, 1181–5.
Thiers, G., Maes, R., van Lierde, R., et al. (1979). Surveillance van besmettelijke ziekten door een net van peilpraktij ken. Tijdschrifti voor Geneeskunde, 12, 781.
Thompson, N.P., Montgomery, S.M., Pounder, R.E., and Wakefield, A.J. (1995). Is measles vaccination a risk factor for inflammatory bowel disease? Lancet, 345, 1071–4.
Tillett, H.E. and Spencer, I.L. (1982). Influenza surveillance in England and Wales using routine statistics. Journal of Hygiene, 88, 83–94.
Webster R.G., Bean W.J., Gorman O.T, Chambers T.M., and Kawaoka Y. (1992). Evolution and ecology of influenza A viruses. Microbiological Reviews, 56, 152–79.
White, C.C., Koplan, J.P., and Orenstein, W.A. (1985). Benefits, risks and costs of immunization for measles, mumps and rubella. American Journal of Public Health, 75, 739–44.
White, L.R., Sever, J.L., and Alepa, F.P. (1969). Maternal and congenital rubella before 1964: frequency, clinical features, and search for isoimmune phenomena. Journal of Pediatrics, 74, 198–207.
Wilkinson, P., Tylden-Pattenson, L., and Gould, J. (1979). Professional attitudes towards vaccination and immunization within the Leeds Area Health Authority. Public Health, London, 93, 11–15.
Witte, J.J. and Axnick, N.W. (1975). The benefits from 10 years of measles immunization in the United States. Public Health Report, 90, 205–7.
Wright, P.F, Kim-Farley, R.J., de Quadros, C.A., et al. (1991). Strategies for the global eradication of poliomyelitis by the year 2000. New England Journal of Medicine, 325, 1774–9.
Yeager, A.S., Davis, J.H., Ross, L.A., et al. (1977). Measles immunization. Successes and failures. Journal of the American Medical Association, 237, 347–51.