The design of an effective surveillance system for animal brucellosis in a region or country depends on many factors, as discussed below.
In this document, we are primarily concerned with B. abortus, B. melitensis and B. suis infections. In some countries, all three species may be present, while other countries may have only a single species. Alternatively, it may not be known with certainty which Brucella species are present, unless bacteriological investigations have been undertaken.
Traditionally, these estimates have been based on data passively acquired from the results of bacteriological and serological data from:
abortion submissions to diagnostic laboratories,
routine testing of on-farm samples, such as milk or blood,
notifications from veterinarians if brucellosis is reportable to the authorities, and
off-farm sampling from markets or slaughterhouses.
All of these results may be biased. For example, diagnostic samples may be representative of herds close to a laboratory or from larger herds where the owners or veterinarians are motivated to submit samples. Slaughterhouse and market samples are probably not truly representative. Most animals will be free from clinical disease, will be older animals, and also decisions to sell animals depends on many factors, often unrelated to disease considerations.
Therefore active surveillance should be undertaken to provide a more reliable estimate of Brucella infection in a region or country. There are three basic approaches to this task:
Undertake a total (census) testing. This is usually impracticable because of cost.
Carry out random (probability-based) sampling, where both groups and individual animals have an equal chance of being sampled.
Carry out non-random (purposive) sampling of suspected high-risk groups of animals. Again, these are likely to be biased if convenience determines origin of samples, such as from bleeding herds, or only at vaccination sites, or close to veterinary clinics, or from cooperative owners.
Ideally, a random sampling programme should be undertaken to provide statistically reliable estimates of the prevalence of infection. This presupposes that a reliable and current sampling frame is available of villages or herds and flocks. If this is not available, it may be necessary to use alternate methods, such as random sampling based on geographical location.
The actual techniques used might be:
administrative areas (districts),
village or herd sizes,
production systems, or
- ecological conditions.
Example: Assume you wish to estimate the baseline prevalence of brucellosis in a province, governorate or similar administrative unit.
Step 1: Primary Sampling Unit = village, herd or flock.
A random sample generated either from a complete listing, or on a geographical basis using map coordinates if a list is not available.
Sample size: If there is no prior knowledge, assume that 50% of the villages or herds or flocks are infected. From the table below, identify the approximate sample size required to estimate prevalence in a very large (infinite) population with the desired fixed-width confidence limits.
Expected |
LEVEL OF CONFIDENCE |
||||||||
90% Desired Accuracy |
95% Desired Accuracy |
99% Desired Accuracy |
|||||||
10 |
5 |
1 |
10 |
5 |
1 |
10 |
5 |
1 |
|
10% |
24 |
97 |
2 435 |
35 |
138 |
3 457 |
60 |
239 |
5 971 |
20% |
43 |
173 |
4 329 |
61 |
246 |
6 147 |
106 |
425 |
10 616 |
30% |
57 |
227 |
5 682 |
81 |
323 |
8 067 |
139 |
557 |
13 933 |
40% |
65 |
260 |
6 494 |
92 |
369 |
9 220 |
159 |
637 |
15 923 |
50% |
68 |
271 |
6 764 |
96 |
384 |
9 604 |
166 |
663 |
16 587 |
60% |
65 |
260 |
6 494 |
92 |
369 |
9 220 |
159 |
637 |
15 923 |
70% |
57 |
227 |
5 682 |
81 |
323 |
8 067 |
139 |
557 |
13 933 |
80% |
43 |
173 |
4 329 |
61 |
246 |
6 147 |
106 |
425 |
10 616 |
90% |
24 |
97 |
2 435 |
35 |
138 |
3 457 |
60 |
239 |
5 971 |
From the above table, if the expected collective prevalence is 50%, then a sample of 96 villages or herds or flocks would be needed for 95% confidence at ±10% desired accuracy. For great accuracy, say ±1%, then the sample size increases considerably, to 9 604.
When sampling from a finite population of size N, an adjustment can be made to account for this using the formula:
1/n = 1/nX + 1/N
where nX is the sample size calculated above.
Using the above example, assume the population of villages or herds (N) was 1 150
Then 1/n = 1/96 + 1/1150, so n = 89.
Thus testing of 89 villages or herds would be sufficient.
Step 2: Secondary Sampling Units = individual animals
Obviously in the case of Brucellosis these would be sexually mature male and female animals, as we are attempting to detecting the number of infected villages or herds, i.e. those with at least one infected animal.
From the following table, assume that the expected individual prevalence within the village or herds or flocks is 15%, and the desired confidence level is 95%. Sample sizes for varying population sizes are as follows*:
Eligible animals |
Sample size |
10 |
10 |
50 |
16 |
100 |
17 |
500 |
19 |
1 000 |
19 |
* The reader is referred to standard epidemiological and statistical texts on different prevalence rates, desired confidence levels and sample size.
Additional villages or herds may be added to compensate for any refusals. All selections must be random, and the final sampling number will be calculated for each strata.
From this exercise it should be possible to estimate:
the collective prevalence, i.e. number of infected villages or herds or flocks in the region, and
the approximate levels of individual animal infection within villages or herds or flocks.
This information can then provide the basic framework needed to establish a surveillance programme to support control or eradication.
Any surveillance system for brucellosis must include a set of specific definitions that clearly describe the terms used so that there is little room for doubt. These could include any or all of the following:
Traditionally the infection status of individual animals has been measured in brucellosis surveillance, but, as stressed earlier, rather than being an individual animal problem, brucellosis is a herd, flock, village or regional epidemiological problem. Therefore, group data are a more accurate measurement of progress or lack thereof. For example, if the individual animal prevalence is gradually decreasing, but the herd or flock prevalence is increasing, then obviously the control programme should be investigated to determine the reasons for this anomaly.
Usually a herd or flock has been defined on the basis of species, ownership and location, such as All animals of the same species or multiple susceptible species under common ownership or supervision that are a group on one or more parts of any single premise (lot, farm, ranch, etc.) or a group of animals that is maintained separate from other animals by an approved fence or natural barrier.
These definitions require modification if animals are on two or more premises that are geographically separated but with interchange of animals between them, or if the animals have contact with animals from different premises. In other situations, animals of the same species may commingle on community pastures, or as part of a village flock or herd - a situation probably best described using geographical coordinates, i.e. by mapping.
Where there is a village system or extensive nomadic or transhumant systems, it becomes more difficult to define the observational unit unless sampling is confined to a specific time of the year where location is reasonably certain. Mixing of small ruminants (sheep and goats) is very common in many countries that rely on extensive grazing. Intensively managed herds or flocks may be self-contained and essentially closed to all other livestock except via the introduction of semen or embryos.
While latent infections (up to 10% of animals born to infected dams) occur in brucellosis, the clinical disease is confined to sexually mature animals. Therefore, age ranges and sex status for surveillance purposes should be clearly defined, such as the following:
Cattle: Test eligible animals, including unvaccinated cattle of 6 months of age and older, and official calfhood B. abortus vaccinates above 18-20 months of age. Note that some countries exclude spayed and castrated animals from testing.
For example, all cattle in a known infected herd or that have been in contact with known Brucella reactors in a market for at least 24 hours.
Freedom may be defined on the basis of time and specific areas, as well as on the history of successful elimination of infected animals or herds or flocks.
These would be dairy, beef, confinement, etc.
This would include vaccinated and reactor animals
Great variation exists among livestock production systems worldwide, from the landless (total confinement) intensive systems of dairy cattle to the extensive husbandry of mixed species grazing with very low animal concentrations per unit area. Obviously, the type of system will affect the rate of spread of infection both within and between villages, herds and flocks. In the early stage of test and slaughter phases, on-farm village sampling is preferred, particularly as the owner can be actively involved in education.
Later, as the prevalence decreases, off-farm sampling at markets or abattoirs is generally more cost-effective, providing that ownership or identification is maintained.
In most livestock, parturition is seasonal, and knowledge of these patterns is important in determining probable times for the occurrence of abortions. Migratory herds or flocks may be more easily located during lambing or kidding periods.
The sale of animals for meat depends on many factors and is usually not random throughout the year. In some cases, long periods may occur with very few animals sold, so a slaughter-based surveillance programme could be inefficient. Markets, especially if terminal (i.e. animals destined for slaughter), are a very useful public event in the life of an animal, when it is accessible for sampling. If movement permits are required, this event can also used for blood sampling.
Testing of bulk milk for brucellosis is a valuable screening test in cattle, and samples may be collected either on-farm where tanker collection is undertaken or at a milk plant where producers bring their milk for sale on a regular basis. Routine milk quality examination samples may also be used for brucellosis testing.
In summary, a careful study of all the livestock systems in relation to brucellosis epidemiology should be undertaken prior to commencing surveillance, facilitating the determination of the most convenient and economical sampling sites.
Most countries have an established system for collecting livestock data, varying from complete census at specified intervals to intermittent sampling programmes. Periodic census data may also be updated by estimates between censuses. As parturitions are often seasonal, the time of the year when sampling occurs should be specified.
Ideally, the number of flocks and herds and their distribution should be obtained.
However, given the dynamics of livestock production, any statistical data soon becomes outdated. Also, if livestock data is related to any taxation system, there will probably be an under-reporting bias. Therefore it is recommended that those responsible for surveillance use complementary information sources, such as district veterinary lists, farmer or cooperative organizations memberships, and even aerial photography, to ensure that any sampling programme is as complete as possible. In some countries, all herds and flocks are registered with the Veterinary Department or another government agency. Where there is multiple ownership of herds o flocks, it may be necessary to keep two registers - one for the main owner who has more than one flock, and another for the direct owner who has only one flock. For grazing purposes, nomadic or transhumant flocks may be required to get permission, and this is given only to those owners that have properly vaccinated and registered their animals.
Many different systems of animal and herd identification are used for disease control and surveillance purposes. For brucellosis surveillance, the minimum should be a herd or flock identifier such as permanent ear or tail tags, ear notches, tattoos or branding. If individual animal identifiers are used, information on their vaccination status should also be included. In some countries, temporary identifiers are placed on animals prior to marketing or slaughter to enable trace-back to the herd of origin for positive test reactors. There is no perfect system of identification as losses occur by accident or intent - nevertheless, a reliable herd or flock identification system is integral to any surveillance system, especially where trace-back to the herd or flock of origin is to be attempted.
In the control and eventual eradication of brucellosis, there are generally four overlapping phases:
The choice of sampling and types of herds or flocks to monitor will obviously depend on the phase of brucellosis control. For example, once the prevalence of infected herds has been reduced to a low level, it is usually uneconomic to continue testing all eligible animals and surveillance can focus on problem herds, abortion incidents, herds adjacent to known infected herds and off-farm testing, such as in markets and slaughterhouses.
Two types of laboratory support are needed for Brucellosis control and surveillance: bacteriological testing, and serological testing.
Bacteriological Tests
Appropriate facilities are needed to isolate and identify all suspect Brucella spp. from abortion materials (foetal stomach contents and cotyledons), milk and vaginal discharges, as well as tissues from slaughtered reactor animals, such as supramammary lymph nodes. Ideally, isolates of Brucella spp. should be fingerprinted by biotyping. Periodically, isolates should be confirmed by submission to a WHO/FAO Collaborating Centre. As laboratory exposures to Brucella spp. have occurred frequently, any laboratory to be used for isolation of brucellae should have a primary biohazard containment facility, to minimize the risk of human infection.
Serological Tests
Many serological tests for brucellosis have been developed, or are under development. However, for serological surveillance to be successful, it is advisable to concentrate on a few tests only, and ensure that these are quality controlled and can be carried out with the facilities available. Given that no test is both 100% sensitive and specific, it is not advisable to use a battery test approach in the mistaken belief that if enough tests are done the results will become clear. Rather, what is needed is a simple and clearly defined testing strategy with defined endpoints, and a rigorous approach to borderline cases, which takes into account the epidemiological features of brucellosis. Note that only certain tests for surveillance purposes are recognized as official by the OIE Manual of Standard Diagnostic Tests and Vaccines (OIE, 2000). If sensitive and specific enough, any country may, however, declare other tests used for diagnosis, control and surveillance purposes in their programmes.
When individual animals are tested to ascertain if the herd is infected, the number of animals tested and the critical number of reactors used to decide the health status of the herd becomes very important in influencing the herd level sensitivity and specificity. If the test specificity is less than 100%, then as the number of animals tested increases the probability of at least one false-positive animal increases, and so the herd specificity decreases. The herd sensitivity, herd negative predictive value and herd apparent prevalence increase directly with the number of animals tested, but the herd positive predictive value decreases. Herd sensitivity can be increased by using a test that is less than 100% specific. These features should be borne in mind when interpreting the natural history of brucellosis, and particularly as it is recognized that larger herds are more likely to be infected than smaller herds.
Serological tests can be divided broadly into two groups:
Screening tests used in the field clinics or in regional laboratories, such as the Rose Bengal or buffered plate agglutination. The Rose Bengal test has a very high sensitivity to ensure that infected animals are not missed. The milk ring test is also an excellent screening test for dairy cattle. Indirect ELISA tests are also being used to screen milk and serum.
Confirmatory tests used in a central or regional laboratory, such as competitive ELISA, immunodiffusion or complement fixation tests, are very useful in distinguishing vaccinal antibody responses from those induced by field infections.
It is important to note that, during the course of a brucellosis programme, testing strategies will change. For example, when the prevalence of infection is high, a test of adequate sensitivity but high specificity is desirable to detect most of the truly diseased animals and herds, and to minimize the number of false-positive reactors. In contrast, as the prevalence decreases, a sufficiently specific but highly sensitive test is recommended. It is thus important to decide how to classify positive animals, such as single reactors in an otherwise negative herd. This may be the first sign of a herd breakdown, or it may be a false positive of no importance. In some countries, the problem of false-positive serological reactions due to cross-reactive bacteria (e.g. Yersinia enterocolitica 0:9) has complicated the eradication of brucellosis.
A number of commercial kit tests are also available and may be particularly useful for confirmatory purposes, but their costs often preclude their use in large surveillance programmes. Automation of some tests may also allow for economies of scale.
Banking of a representative sample of sera in a deep freeze is strongly recommended for retrospective investigations of problem herds.
The integrity and accuracy of any surveillance system is dependent on how data is recorded in the field and laboratory, how it is transcribed and summarized, and finally how it is interpreted. Computerized databases are now gradually replacing manual recording systems. A variety of paper records are required for primary data entry, including:
district, province, region, day of visit, owner, address, latitude and longitude;
the species, number of animals, breed, production system (housing or grazing), and breeding policy;
fertility and abortion rates, and other signs; and
history of brucellosis, control measures and vaccinations.
A herd or locality identifier should be included.
Note: The above forms (i-iv) are available from the Animal Production and Health Section of the Joint FAO/IAEA Programme for Sero-Monitoring of Brucellosis, in the form of a software database.
Other forms may be developed depending on individual requirements. In terms of the amount of information to be recorded, a balance needs to be made between what is essential for epidemiological and surveillance analyses (need to know) and what could be termed nice to know, i.e. of minimal value for decision-makers. Regular validity checks should be carried out on a percentage of records (5%) to determine obvious errors and assess the extent of missing data.
It is important to distinguish between zero reporting and non-reporting of zero incidences. The former indicates that the reporting office is active, whereas the latter cannot be distinguished from failure to conduct surveillance or report.
Other computer software that could be used to develop a brucellosis database includes:
TAD info- FAO software Java Version. Website: http://www.fao.org/ag/AGA/AGAH/EMPRES/tadinfo2/e_tadinf.htm
Epi Info (with or without Epi Map). This is available in the USA from:
USD, Inc.,
2075-A West Park Place,
Stone Mountain, GA 30087
Telephone (770) 469 4098
Fax (504) 469 0681
E-mail: [email protected]
Website: http://www.usd-inc.com
It can also be downloaded from: www.cdc.gov/ > Publications and Products > Click on Epi Info and Epi Map. Instructions are provided for downloading. Free Technical Support for these Programmes can be obtained by e-mail at [email protected]. Epi Info is now available in 12 non-English languages.
Active Surveillance for Livestock Diseases - Practical Techniques for Developing Countries. A manual and associated software is available from Dr Angus Cameron, Australia [email protected],
Database for the Surveillance of Rinderpest and Sero-monitoring of Brucellosis, Joint FAO/IAEA Programme, obtainable from IAEA, Vienna, Austria. Website: http://www.fao.org/ag/age/d3/index.html
Note that computerized mapping programmes are also now used to identify brucellosis-infected and brucellosis-free zones or areas in countries. These are especially valuable if there is a unique farm identifier.
Performance, Diagnostic and Resource (Workload) indicators, as described earlier, will vary with both the species (bovine, ovine, caprine or porcine) affected and also the phase of the brucellosis programme, i.e. high or unknown prevalence; mass vaccination; test and removal, segregation or slaughter; and freedom.
Performance indicators
Ideally these are time-delimited rates with a numerator (infected animals, herds, flocks or villages or other administrative unit) and a denominator (at risk in the above categories). As mentioned, incidence rates based on groups of animals are the most sensitive indicators of success or failure of a programme. Alternatively, the ratio of newly identified herds for the current year compared to the previous year can be used to obtain the percentage reduction (or increase), calculated as:
Current year -previous year
Previous
year
Similar prevalence rates for all known infected herds can be calculated.
Diagnostic indicators
Examples include:
Number of complete herd tests at monthly intervals needed to clear herds of infection.
Mean quarantined period of a herd or flock as an effectiveness of control.
Relative sources of sero-positive animals as leading to infected herds or flocks.
Efficiency of trace-back procedures.
Number of epidemiological investigation carried out.
Number of culture-positive animals in relation to the number sero-positive (slaughter examinations).
Number of serological tests in relation to their classification of infected herds. The more serological tests, the higher the possibility of a better quality diagnostic system.
Number of animals (or herds) vaccinated as a fraction of eligible animals (or herds).
Resource indicators
These can reflect the manager used, the budget spent or the resources used. Examples include:
Costs per animal for vaccination.
Costs for serology, bacteriology, etc., on a per-test basis.
Costs for epidemiological investigations.
Surveillance data can be used in several general ways for decision-making in brucellosis control and eradication programmes, including to
investigate sources of infection for individual herds or flocks;
use aggregated descriptive data to monitor trends over time and space to detect problems as they occur; and identify risk factors using herds or flocks in a case-control study format,
and these are discussed below.
Investigation of sources of infection for individual herds or flocks
All newly identified infected (or re-infected) herds or flocks should be carefully investigated using a standard form to record information such as:
Owners full address or location, and date first positive tests recorded.
Reason for testing, such as on- or off-farm, diagnostic, movement; area test; or following trace backs from other infected herds, adjacent or contact herds on common grazing.
Humans known to be infected.
Clinical signs present, such as abortion, retained placenta, difficult breeding, or reduced milk production.
Percentage vaccinated and type (young or adults).
Type of operation and total census of animals.
Breeding programme and usual timing of parturitions.
Origin of herd or flock.
Recent movements in or out, or both, associated with an infected herd or flock.
Names of adjacent owners of livestock.
Contact with other susceptible species.
Quarantine (if applicable) requirements.
Probable source of infection and date introduced.
In addition to the origin of reactor animals, other recent purchase or loan animals and animals removed should be recorded as carefully as possible. Finally, all the herds in the immediate area should be visited to determine if commingling is likely to occur with the infected herd(s) and whether they should also be tested.
Biotyping of the Brucella species may be used to assist in identifying potential sources of infection. This includes differentiation of wild from vaccine strains, which occasionally may cause abortion.
From this information it should be possible to determine if the source for the infected herd was either endemic infection - i.e. a prior history of brucellosis - or a recent infection by introduction of infected animal(s) or by direct contact with infected animals.
Use of aggregated descriptive data to monitor trends to detect problems as they occur
Aggregated herd and flock data should be examined for the whole country or regions within the country to determine trends over time (monthly and annually), and also by place, such as districts, provinces or governorates. Such information can be displayed by graphically using maps, graphs or tables using the appropriate software.
It is also useful to compare the success rate of the various surveillance sources, particularly in relation to costs. For example, routine on-farm testing of all herds or flocks compared with off-farm testing at markets, slaughter or through bulk milk testing.
Identification of risk factors, using herds or flocks in a case-control study format
Case-control studies can be carried out, using as a case definition an infected flock or herd with more than 5% of animals infected, with a known Brucella-free flock or herd as a control. Cases and controls should not be matched, except perhaps by region. From these studies, odds or ratios (or approximate relative risks) can be calculated to determine the effect of herd size, type of management, and vaccination status and production system as potential risk factors. Thus, high risk herds can be targeted for additional investigations.
The costs of brucellosis control or eradication programmes are high, and involve considerable government resources, as well as contributions from the livestock industries. Without strong political support, programmes may be under-funded, particularly in the final stages, when the number of infected animals or herds detected is very low in relation to the inputs needed.
Legal factors are important in their effect on the enactment and enforcement of laws and regulations, such as for identification or compensation for infected animals. These may act to either encourage or discourage the detection and reporting of the health-related events. Ideally, representatives from livestock industries should have input into the development of government regulations so that they are not only scientifically defensible, but also perceived to be user friendly, and thus more likely to ensure compliance, and, ultimately, more accurate surveillance.
It is an administrative responsibility to ensure that any surveillance programme is adequately staffed with veterinarians, technicians and support staff so as to be able to carry out the programme within the planned time frame and budget. They should also be well trained and motivated to carry out sampling exactly as required by the epidemiologists.
Ideally, the veterinarian(s) responsible for the surveillance programme should have had advanced training in epidemiology and biostatistics, and be familiar with computer programs.
The surveillance activities preferably should be administratively separate from the actual control or eradication activities within government veterinary services. In effect, this acts to ensure that the epidemiologists present surveillance information in an unbiased format to the decision-makers. Budgets for surveillance are limited, and administrators must balance acceptable costs and acceptable risks.
Cultural and social pressures often discourage reporting of animal diseases. In some societies, it may be necessary to convince the leader or village headman, for example, to facilitate sampling of animals. To overcome these barriers, educational efforts must be combined with changes in the reward system to provide some positive incentives for those persons or groups involved in farm surveillance. These rewards should vary to fit their desires and perceptions. For example, some may be rewarded by the timely investigation of abortion outbreaks and prompt return of results with interpretation and intervention whether the abortion is due to brucellosis or not. Others may need more tangible rewards to ensure cooperation, such as free medication for their livestock. Only through positive motivation will it be possible to obtain a high level of detection, accurate classification and timely reporting of brucellosis health-related events.
If current surveillance information shows that a brucellosis control programme is not achieving its desired objectives despite being technically sound, there may be significant problems in knowledge, attitudes or practices among either the public or the livestock producers. This is especially likely to be a problem in the final stages of an eradication campaign, when livestock producers may not have experienced brucellosis and no important economic losses due to the disease (abortion), and thus question its relevance. There are well-established techniques available using Participatory Rural Appraisal methods to determine knowledge, attitudes and practices associated with brucellosis in livestock owning communities. This is very clearly a legitimate component of brucellosis surveillance.