Veterinary Epidemiology Principles

The document initially highlights a significant change in approach within domestic animal industries. The focus has shifted from examining individual animals to considering animal populations as aggregates. This shift is intrinsically linked to the increasing integration of epidemiology into health maintenance programs. The text suggests that as population-oriented health programs become more common in livestock production systems, the importance of epidemiology will only continue to grow. This necessitates a broader perspective, moving beyond individual animal care to encompass the overall health of the herd or population. The authors hope the methodologies presented will encourage greater private practitioner involvement in formal health management programs, both independently and collaboratively with dedicated epidemiology units. This collaborative approach is deemed crucial for effective disease control strategies and improving the efficiency of health management across the board.

This section introduces the core concept of disease determinants—factors that influence health and disease. The text emphasizes that while determinants are commonly referred to as causes, in epidemiology, a determinant is defined as any factor whose alteration produces a change in disease frequency or characteristics. This highlights the complexity of disease causation, arguing against the notion of a single cause for most diseases. The document distinguishes between host factors (intrinsic, such as age, breed, and sex) and external determinants. Putative causes are further described as exposures or risk factors, or independent/predictor/explanatory variables—all terms referring to factors suspected of influencing the outcome (health status or disease occurrence). The outcome is conversely defined as the dependent variable—the presumed effect, often measured as health (productivity) or disease incidence. This section builds a foundational understanding of the multifaceted nature of disease etiology, setting the stage for more in-depth analysis of causal relationships.

The text establishes a clear distinction between clinical and subclinical disease. Clinical disease is defined as a state of bodily dysfunction detectable by the senses, while subclinical disease is a functional or anatomical abnormality only detectable via laboratory tests. Though subclinical disease may be less severe for an individual animal, its high prevalence often makes it more significant at the population level. The document stresses that, regardless of the primary cause(s), the number of subclinically diseased animals typically far exceeds the number of clinically diseased animals. A critical distinction is made between infection and disease; many animals infected with disease agents may show no clinical signs. The importance of this distinction in understanding disease dynamics and implementing effective control strategies is emphasized. The concept that most animal populations contain varying proportions of healthy, subclinically diseased, and clinically diseased individuals, with these proportions subject to change over time, is highlighted. This dynamic perspective is crucial for understanding and predicting disease trends and planning effective interventions.

The predictability of disease occurrence in populations forms the basis for many epidemiological field studies. Veterinarians implicitly and explicitly use this knowledge; clinicians implicitly utilize the predictability of diseases associated with specific life stages (e.g., milk fever in cows near parturition). Epidemiologists explicitly utilize this knowledge to identify risk factors associated with certain conditions (e.g., the higher likelihood of feline urologic syndrome in indoor, castrated cats fed dry food). This inherent predictability encourages investigation into why a disease occurs under specific circumstances. For instance, the text questions why wildlife rabies seems more prevalent in urbanized areas than in isolated rural settings. This emphasizes the importance of understanding environmental, behavioral, and other contributing factors in the broader context of disease dynamics and distribution patterns.

The text champions the use of observational studies in veterinary epidemiology, particularly when controlled experiments are impractical. These studies, characterized by the researcher's role as an observer rather than manipulator, leverage naturally occurring situations. A key example cited is John Snow's investigation of cholera in London. Snow ingeniously used a screening test to identify the water source for each household, demonstrating a strong association between water from Southwark and Vauxhall companies and higher cholera rates compared to Lambeth's supply. This insightful study highlights the power of careful observation and data collection in uncovering causal relationships, even without direct experimental manipulation. The success of this observational study underscores the value of well-designed observational approaches in epidemiological research, especially when practical constraints limit the feasibility of controlled experiments. The methodology employed by Snow also illustrates the importance of choosing appropriate sampling units, emphasizing that if the unit of concern is the individual, then individual animals should be the sampling unit to properly detect a direct cause of disease.

In contrast to observational studies, experimental approaches, including controlled experiments and field trials, are discussed. While laboratory experiments offer precise control over treatment and challenge, field trials aim to assess the effectiveness of products and programs under realistic conditions. The text emphasizes the importance of formal random allocation of treatment groups in true experiments to minimize bias and ensure valid results. Quasi-experiments, where the investigator personally assigns treatments, are acknowledged but considered less likely to yield valid results, especially when aiming to prove a specific point rather than solve a biological problem. The importance of randomization is highlighted – it allows the calculation of the probability that observed differences are due to chance, rather than guaranteeing identical groups. The section emphasizes the need for both observational and experimental approaches in veterinary epidemiology to achieve a more comprehensive understanding of disease processes. A comparison is made between the evidence derived from observational versus experimental studies, showing that while experimental studies offer stronger evidence, observational studies are simpler to perform.

This section addresses the challenges of confounding variables in observational studies. Confounding occurs when a third factor is associated with both the exposure and outcome of interest, leading to biased estimations. An example is provided involving a fictitious study of staphylococci and mastitis in dairy cows, where streptococcal organisms act as a confounding variable. Ignoring streptococci in the analysis leads to a biased relative risk. Methods for addressing confounding, such as matching, are discussed. Matching involves selecting exposed and unexposed groups to balance the distribution of confounding variables. This strategy helps to ensure the comparability of groups and reduce bias in the analysis. The discussion emphasizes the importance of careful consideration and control of confounding variables to ensure valid conclusions in epidemiological studies, highlighting the potential of confounding to either exaggerate or obscure true relationships.

This section emphasizes the importance of accurately measuring disease frequency in animal populations. It contrasts the use of incidence rates and prevalence proportions with less informative measures like case counts. Incidence rates and prevalence proportions offer estimates of the risk (probability) of disease occurrence at different levels of host factors (e.g., comparing males versus females, young versus old animals, or different breeds). Case counts, on the other hand, are influenced by both the risk of disease and the size of the population subgroup, making them less suitable for direct inferences about risk. The text cautions against drawing conclusions about disease risk solely based on case counts without adjusting for the population at risk. An example of this is provided with mastitis, where the higher number of cases in a particular age group of cows does not automatically indicate increased risk unless compared to the age distribution in the source population. This section highlights the critical need for precise measurement of disease occurrence to accurately assess risk and inform effective disease control strategies.

The section discusses the importance of selecting appropriate parameters for measuring animal productivity, especially considering their economic implications. The example of milk production in dairy cows is used to illustrate the complexities. Absolute measures like total kilograms of milk produced in a lactation (kg tot) and kilograms produced in a 305-day period (kg 305) are introduced, with the kg 305 preferred for removing variability due to differences in days-in-milk. However, age and calving season can still influence kg 305. To account for these confounding factors, the breed class average for milk production (BCM) index is presented as a standardized measure. This standardized approach facilitates comparisons between groups of animals of different ages and calving seasons. The text emphasizes the need to choose parameters that are not only good measures of production but also serve as valuable economic indicators for decision-making in animal health management. A hierarchy of parameters is recommended to effectively monitor production changes in health management programs, emphasizing the importance of monitoring both mean and standard deviation values.

This section addresses the critical issue of detecting subclinical disease in animal populations. It highlights that a major source of economic loss in domestic animals comes from the impact of hidden or subclinical diseases. Subclinical mastitis is given as a prime example: though mild and often inapparent, its high prevalence significantly impacts dairy herd productivity more than the sporadic, dramatic clinical cases. The section emphasizes the need for screening tests to detect these hidden infections. Screening, defined as applying tests to apparently healthy animals to detect infections or subclinical disease, is presented as a crucial tool for understanding disease processes and the role of various agents in syndromes like pneumonia or gastroenteritis. The text argues that focusing on how infections occur and persist even in the absence of clinical disease offers a more effective approach to disease prevention than solely relying on diseased animals as study models. This reflects a shift towards a more proactive and preventative approach to animal health management.

This section details strategies for sampling animal populations to detect disease. The text acknowledges the limitations of complete examination of large populations (e.g., examining the snout of every pig in a 5000-pig operation). It argues, however, that sampling can provide valid insights because infectious diseases tend to spread within herds. Thus, if disease is present, it's likely to affect more than one animal. The sampling strategy is designed to detect disease if more than a specified number or percentage of animals are affected. The specified number or percentage should be based on the disease's biology and potentially on prior testing data. An example is provided with bovine tuberculosis, where previous data suggesting a prevalence of 5-10% in infected herds could inform sample size calculations for new surveys. This section illustrates the importance of a targeted sampling approach to maximize the efficiency and effectiveness of disease detection within animal populations.

This section introduces descriptive epidemiology, focusing on the careful documentation of disease occurrence and patterns. The example of Contagious Bovine Pleuropneumonia (CBPP) in the United States illustrates this approach. Initially, explanations for CBPP outbreaks relied on vague notions of contagion. However, through meticulous recording of cases and outbreaks, investigators determined that imported or purchased cattle were the primary source of infection in nearly all instances. Further experiments confirmed the contagious nature of CBPP, ruling out spontaneous generation. Observing that the disease spread more rapidly and severely during summer months proved helpful in the eradication program. The successful eradication of CBPP in the US by 1892, even before the causative agent (a mycoplasma) was identified, highlights the power of descriptive epidemiology in disease control. The successful CBPP eradication campaign, along with John Snow's cholera work, shows that effective disease control is possible even without a complete understanding of the disease's etiology or pathogenesis, provided sufficient knowledge of its natural history is available. Understanding the natural history often reveals weak links in the causal chain that can be targeted for intervention.

This section transitions to analytic epidemiology, focusing on identifying associations between factors and disease. It emphasizes that the epidemiologic use of 'association' differs from common usage. Two events occurring together in the same individual are not necessarily epidemiologically associated. To establish an epidemiologic association, the events must occur together more or less frequently than expected by chance alone, necessitating a comparison group (e.g., comparing diseased animals to non-diseased animals). The text uses the example of Haemophilus somnus isolation from cattle with pneumonia. Simply finding the organism in diseased lungs doesn't prove association; a comparison group of cattle without pneumonia is needed to determine if the organism is more or less frequent in diseased animals. Statistical tests are required to evaluate the likelihood that observed associations are due to chance. This section underscores the critical difference between simple co-occurrence and epidemiologically significant associations, emphasizing the need for robust statistical analysis and proper control groups in analytic epidemiological studies. This lays the groundwork for more complex analytical techniques to establish causal relationships.

The section provides several practical examples illustrating descriptive and analytic epidemiological principles. The Contagious Bovine Pleuropneumonia (CBPP) eradication program in the United States is detailed, noting the timeline of its spread (reaching Illinois, Kentucky, and Missouri by 1886), the resulting meat export embargo, and the eventual success of the eradication program by 1892. This successful program, despite a lack of complete etiological understanding at the time, highlights the importance of a strong public health response based on the available knowledge of the disease's natural history. The example of John Snow's cholera study in London is revisited, emphasizing the effectiveness of his household-level analysis to pinpoint contaminated water as the source of the outbreak. The contrasting approaches and outcomes of these real-world case studies demonstrate the effectiveness of both descriptive and analytic techniques, and the importance of understanding disease spread in different contexts.

This section tackles the central challenge of establishing causality in observational field studies. The difficulty lies in the investigator's inability to ensure that other factors didn't contribute to the observed outcome. While laboratory experiments allow for greater control, establishing causality in field studies is more complex due to the numerous known and unknown factors influencing disease occurrence. The text contrasts this with well-designed laboratory experiments, where a statistically significant difference in disease rates between exposed and unexposed groups strongly suggests a cause-and-effect relationship (Method of Difference). In field trials, even with controlled treatment allocation, unknown factors can still influence outcomes, weakening the certainty of causal inference. Therefore, additional corroborating evidence from independent studies is often necessary. Observational studies are even more susceptible to bias due to multiple known and unknown factors. The inherent limitations of observational studies are acknowledged, but the text emphasizes the need for more complex study designs to compensate and provide additional guidelines for drawing causal inferences from observational data.

The concept of confounding variables is introduced, illustrating how these variables, associated with both the exposure and outcome, can bias study results. A fictitious example involving staphylococci and mastitis in dairy cows, with streptococcal organisms as a confounding factor, demonstrates this issue. Ignoring the confounding variable (streptococci) leads to an overestimation of the relative risk associated with staphylococci. The section discusses strategies for reducing bias caused by confounding, primarily matching. In cohort studies, matching involves selecting unexposed groups that mirror the exposed group's distribution of confounding variables. A similar process is employed in case-control studies. The text points out that while matching can minimize bias, it doesn't eliminate it entirely and may even reduce the statistical power of the analysis if improperly applied. The importance of careful selection and consideration of variables used for matching is stressed to ensure robust and reliable study results.

This section defines direct and indirect causes. A factor is a direct cause only if there is no known intervening variable between the factor and the disease, and both variables are measured at the same level of organization. All other causes are indirect. The text argues that while researchers often seek direct causes, controlling disease might be easier by manipulating indirect causes. An example is given where, although pathogens are direct causes, controlling housing or management might be a more effective preventive measure. The distinction is also presented in the context of evolving knowledge: the lack of citrus fruits was once considered a direct cause of scurvy, but after the discovery of vitamin C, it became an indirect cause. The identification of a direct cause opened up new avenues for prevention (e.g., synthetic ascorbic acid), but the importance of citrus fruits in preventing scurvy wasn't diminished. This highlights that determining the most direct cause is valuable but not always the most practical approach for disease control.

This section introduces the core principle of experimental methods: the planned comparison of outcomes in groups receiving different treatments. Examples of treatment levels include vaccine versus no vaccine, or different therapeutic drugs. Outcomes might include disease rates or animal productivity. The goal in designing field trials is to ensure treatment group comparability and minimize experimental error, thereby identifying treatment effects efficiently. Comparability relies on the treatment allocation method and group management during the trial. At the study's end, statistical tests determine the likelihood that observed differences between treatment groups arose from chance variation. The text highlights that the treatment in a field trial could range from a single drug to a comprehensive program encompassing multiple interventions. The section also stresses the importance of ensuring the comparability of the treatment groups to draw valid conclusions from the study results.

The document outlines several experimental designs commonly used in veterinary field trials: completely randomized, randomized complete block, crossover, factorial, and split-plot designs. While it doesn't delve into the specifics of each, it highlights their general applications and advantages, particularly the ability of factorial and split-plot designs to assess interactions between multiple treatments. The split-plot design is especially useful when one treatment can only be applied to aggregates (e.g., adding antimicrobials to the water supply of a pen), while the other can be allocated individually (e.g., vaccinating individual animals within that pen). The text strongly suggests seeking the advice of a statistician during the planning stages of any field trial to ensure the chosen design is appropriate for the research question and the resources available. The section emphasizes that the choice of experimental design has significant implications for both the execution and the interpretation of the study results.

This section addresses biological factors that can impact the design of vaccine and therapeutic trials. Herd immunity, where a resistant majority protects a susceptible minority, can mask treatment effects, leading to false negatives. Similarly, applying treatment to only a small proportion of a herd increases the challenge to the untreated animals, potentially leading to inaccurate conclusions. To overcome these issues, the use of physically separated experimental units (e.g., randomizing herds rather than individual animals) is recommended, particularly when testing anthelmintics, where untreated animals may contaminate the environment and increase the challenge for treated animals. The importance of identifying the correct experimental unit is emphasized. The experimental unit is defined as the smallest independent grouping of elements randomized to treatment groups. Errors in identifying the experimental unit can lead to inaccurate interpretation of results, rendering the study unable to generate valid statistical conclusions. Examples of this include using herds rather than individual animals or housing individuals together within a pen, where the pen, not the individual, becomes the effective experimental unit.