Designing Longitudinal Studies in Veterinary Populations

By Dr. Zubair Khalid, DVM, MS, PhD · 2026-08-11

Key Takeaways

• Longitudinal studies in veterinary populations are designed to track changes in biomarkers, disease incidence, or survival over time by repeatedly observing the same animals, thereby distinguishing within-individual change from between-individual variation, a critical advantage over cross-sectional designs that conflate age, period, and cohort effects.
• Prospective cohort studies are ideal for assessing incidence and prognostic factors when exposure requires active measurement, while repeated measures designs are suited for continuous outcomes like body weight or biomarker concentrations, with measurement intervals dictated by the biologically relevant time scale, ranging from days for acute conditions to years for chronic diseases.
• Time-to-event outcomes, such as mortality or disease onset, necessitate survival analysis, which accounts for variable follow-up times and censoring; precise definition of the event, time origin, and handling of competing risks (e.g., death from unrelated causes) are paramount for valid inference.
• Statistical analysis of longitudinal data must account for the correlation between repeated measurements using methods like mixed-effects models for continuous outcomes or generalized estimating equations for non-continuous outcomes, and robust strategies for handling missing data, such as multiple imputation, are essential given inevitable attrition.
• Rigorous protocol design includes standardized measurement instruments with documented repeatability coefficients, meticulous calibration schedules, operator training, and detailed documentation of all data capture and quality assurance procedures to mitigate common failure modes like measurement drift and outcome misclassification.
• Sample size calculations must consider both between- and within-individual variance components, and for time-to-event outcomes, the number of events is the primary driver of power, necessitating over-enrollment to compensate for anticipated loss to follow-up, estimated at 10-20% per year in companion animals and potentially higher in production systems.

Longitudinal studies follow defined animal populations forward through time, recording repeated observations or waiting for clinical events to occur. This article provides a procedural framework for designing such studies in veterinary settings, from conceptual foundation through implementation and analysis. It serves veterinary researchers, graduate students, and clinicians engaged in academic or practice-based research who need to select an appropriate longitudinal architecture, anticipate common failure modes, and plan data collection that will support valid inference.

The central question a longitudinal design answers is one of change or occurrence over time: How does a biomarker progress with disease? What is the incidence rate of a condition in a defined population? Which exposures predict survival? These questions cannot be answered with a single cross-sectional snapshot, because cross-sectional data conflate age effects, cohort effects, and period effects. Longitudinal designs separate these temporal dimensions by observing the same subjects repeatedly, allowing within-individual change to be distinguished from between-individual variation.

At a Glance

Conceptual Foundations of Longitudinal Inference

Longitudinal designs rest on a logic of temporality. Exposure is measured before outcome occurrence, which permits causal ordering that cross-sectional designs cannot establish. The CDC principles of epidemiology in public health practice describe cohort studies as the observational design best suited to establishing incidence and evaluating causal associations, because the investigator observes the sequence from exposure to outcome directly.

Three temporal effects complicate longitudinal inference. Age effects are biological changes that occur with maturation or senescence. Period effects are societal, environmental, or management changes that affect all subjects at a given calendar time. Cohort effects are differences between groups born or enrolled at different times that persist throughout follow-up. A longitudinal design with staggered enrollment and repeated measurement can separate these effects only if the schedule is planned with sufficient spread across both age and calendar time.

Within-Individual and Between-Individual Variation

The statistical power of a longitudinal study derives largely from its repeated measurements. Each subject serves as its own control, so stable between-individual characteriztics such as breed, sex, or genetic background are automatically accounted for in within-individual comparisons. This property makes longitudinal designs efficient for detecting change, but it also creates a requirement: the measurement instrument must be stable across time and across observers. A biomarker assay that drifts between batches, or a clinical scoring system applied differently by different assessors, will produce artefactual change that is indistinguishable from true biological change.

Time Scales and Visit Schedules

The biologically relevant time scale must be specified before enrollment. For chronic diseases with slow progression, such as chronic kidney disease in dogs, the time scale is months to years, and the survival analysis of dogs with CKD in UK veterinary practices demonstrates that follow-up periods of two years or more are needed to capture meaningful outcomes. For acute conditions or fast-growing tumors, the time scale may be days to weeks, as illustrated by the longitudinal MRI assessment of glioma vascular progression in mice, where weekly imaging captured meaningful changes in tumor permeability and vascularity over four weeks.

Prospective Cohort Architecture

The prospective cohort study is the canonical longitudinal design. The investigator defines a population, measures exposures at baseline, and follows subjects forward to observe outcomes. This architecture suits questions about incidence, natural history, and prognostic factors. It is the design of choice when exposure measurement requires active data collection that cannot be reconstructed from records.

Defining the Source Population

The source population must be defined with precision. For companion animal studies, this may be the patient population of one or more veterinary practices, defined by species, age, and presenting complaint. For production animals, it may be a defined herd, flock, or cohort within a management system. For laboratory animals, it is the strain, sex, age, and husbandry conditions specified in the protocol. The WOAH animal health surveillance standards emphasize that the target population and the sampling frame must be explicitly described, because surveillance and research findings are only interpretable in relation to the population from which they were drawn.

Enrollment and Baseline Measurement

Enrollment should occur at a defined biological time origin. This may be birth, weaning, diagnosis, or entry into a management system. Baseline measurements must capture all potential confounders, including demographic variables, clinical history, and environmental or management factors. The baseline period is the only time at which exposure measurement is uncontaminated by outcome knowledge, so it deserves disproportionate planning effort.

Follow-Up and Outcome Ascertainment

Follow-up protocols must specify the visit schedule, the measurements taken at each visit, and the procedures for detecting outcomes that occur between visits. Standardized case definitions should be written before enrollment begins. For survival outcomes, the definition of the event, the time origin, and the handling of competing risks must be explicit. The MSD Veterinary Manual provides species-specific guidance on clinical signs, diagnostic criteria, and disease staging that can anchor outcome definitions in recognized clinical standards.

Repeated Measures Designs

Repeated measures designs are a subset of longitudinal studies in which the same outcome is measured multiple times on each subject, and the primary analytic interest is the trajectory of that outcome over time. These designs are well suited to continuous outcomes such as body weight, biomarker concentrations, imaging parameters, or behavioral scores.

Choosing Measurement Intervals

The interval between measurements must be short enough to capture the biologically relevant rate of change but long enough to be feasible and to avoid measurement burden that drives attrition. For slowly changing outcomes, widely spaced measurements are appropriate. For rapidly changing outcomes, the interval must be shorter than the expected time to meaningful change. The longitudinal study of clinical crown length changes from age 12 to 19 years illustrates a sparse but adequate schedule: measurements at three time points across seven years captured a monotonic process of passive eruption that continued throughout the teenage years.

Measurement Standardization

Repeated measurements amplify the consequences of measurement error. A small systematic drift in a laboratory assay, or a subtle shift in how a clinician applies a scoring system, accumulates across visits and can produce spurious trajectories. Standardization strategies include using a single laboratory for all assays, calibrating instruments before each measurement session, using automated or semi-automated measurement where possible, and training all assessors to a common standard with periodic refresher sessions.

Time-to-Event Outcomes

Survival analysis is the analytic framework for time-to-event outcomes. The outcome is not simply whether an event occurs, but when it occurs, and subjects who do not experience the event during follow-up are censored. This framework accommodates variable follow-up times, which are inevitable in veterinary populations where animals may be sold, moved, lost to follow-up, or die from unrelated causes.

Defining the Event and Time Origin

The event definition must be clinically meaningful and objectively ascertainable. All-cause mortality is unambiguous but may obscure cause-specific effects. Cause-specific mortality requires diagnostic confirmation that may be incomplete in practice. The time origin must be defined identically for all subjects, common choices include birth, enrollment, diagnosis, or treatment initiation. The survival analysis of dogs with chronic kidney disease used diagnosis as the time origin and all-cause mortality as the outcome, with censoring at the end of the study period or loss to follow-up.

Censoring and Competing Risks

Right censoring occurs when a subject is lost to follow-up or the study ends before the event occurs. Informative censoring, where the reason for loss is related to the outcome, biases survival estimates. Competing risks arise when a subject experiences an event that precludes the event of interest, such as death from trauma before the development of a chronic disease. The analytic plan must specify how competing risks will be handled, because standard Kaplan-Meier methods treat competing events as censoring, which overestimates the cumulative incidence of the event of interest.

Longitudinal Data Analysis Principles

Longitudinal data violate the independence assumption of ordinary regression models because repeated measurements from the same subject are correlated. Analytic methods must account for this correlation to produce valid standard errors and hypothesis tests.

Mixed-Effects Models

Linear mixed-effects models accommodate continuous outcomes measured repeatedly. Fixed effects estimate population-average associations, while random effects capture subject-specific deviations from the population average. This framework handles unbalanced data, where subjects have different numbers of measurements, and allows the analyst to model both the mean trajectory and the between-subject variability in trajectories.

Generalized Estimating Equations

Generalized estimating equations provide a population-average approach for non-continuous outcomes, such as binary or count outcomes measured repeatedly. These models are robust to misspecification of the correlation structure and are appropriate when the research question concerns population-average effects instead of subject-specific trajectories.

Handling Missing Data

Missing data are inevitable in longitudinal veterinary studies. Animals die, are sold, move to other practices, or miss scheduled visits. The analytic approach to missing data depends on the mechanism. Missing completely at random occurs when missingness is unrelated to both observed and unobserved values. Missing at random occurs when missingness is related to observed values but not to unobserved values. Missing not at random occurs when missingness is related to the unobserved outcome itself. The longitudinal assessment of taste changes in college-aged males illustrates a common pattern: subjects with complete data differed systematically from those with incomplete data, and the analysis adjusted for baseline characteriztics to reduce bias from differential attrition.

Practical Protocol Design

Selecting the Measurement Instrument

The measurement instrument determines the granularity of change you can detect and the burden you place on the animal and the clinical team. For structural outcomes such as dental crown height or tumor dimensions, direct physical measurement with calibrated instruments remains the reference standard. The longitudinal study of clinical crown length from age 12 to 19 years used digital calipers on serial dental models, a method that yields continuous data with low within-observer variability when the same operator performs all measurements. For physiological outcomes, imaging modalities offer noninvasive repeated assessment. Functional magnetic resonance imaging in awake animals permits brain-wide mapping of stress-related circuits across multiple time points, but it requires substantial investment in animal habituation, anesthesia protocols, and motion correction.

Choose instruments that have published repeatability coefficients in the target species. If the coefficient of variation exceeds the expected biological change, the study will be underpowered regardless of sample size. For field studies, point-of-care analyzers and portable ultrasound are practical but may trade precision for accessibility. Document the instrument model, calibration schedule, and operator training in the study protocol before enrollment begins.

Equipment and Consumable Decisions

The choice between invasive and noninvasive sampling changes the acceptable visit frequency. Serial blood sampling requires venous access planning, particularly in small rodents or neonatal livestock where total blood volume limits sampling frequency. Imaging-based outcomes avoid this constraint but introduce anesthesia risk and scheduling complexity. The dual-bolus perfusion MRI strategy used in a mouse glioma model required two contrast agents administered sequentially, a protocol that demands precise timing and a stable anesthetic plane across repeated sessions.

For each measurement modality, specify:

• Calibration frequency and method
• Consumable lot tracking, particularly for reagents and contrast agents
• Storage and transport conditions for samples
• Maximum allowable time between collection and analysis
• Operator certification requirements

Monitoring Parameters and Their Interpretation

Monitoring serves two distinct purposes: tracking the primary outcome and detecting adverse events or protocol deviations. The table below summarizes common monitoring parameters by outcome domain.

The longitudinal assessment of taste perception in college-aged adults measured suprathreshold intensity at three concentrations across three visits in one academic year, detecting sex-specific changes with weight gain. This design illustrates that the monitoring interval must match the expected rate of biological change. A slowly progressive condition such as chronic kidney disease in dogs requires longer follow-up, and survival analysis with Cox proportional hazards regression is the appropriate analytic framework when the outcome is time to death or euthanasia.

Sample Size and Power Considerations

Variance Components

Longitudinal designs require sample size calculations that account for two variance components: between-individual and within-individual. The within-individual correlation, often expressed as the intraclass correlation coefficient, determines the effective sample size. When repeated measures are highly correlated, each additional visit adds less information than an additional animal. Conversely, when within-individual variability is high, more frequent measurement may be more efficient than increasing enrollment.

For time-to-event outcomes, the number of events, not the number of animals, drives statistical power. Loss to follow-up reduces the event count directly. Assume a loss rate of 10 to 20 percent per year in companion animal populations and higher in production animal systems where culling is common. Inflate enrollment accordingly.

Practical Sample Size Rules

• For mixed-effects models, plan at least 10 to 15 clusters (animals or herds) per covariate in the model.
• For generalized estimating equations, the number of clusters should exceed 30 for valid asymptotic inference.
• For survival analysis, aim for at least 10 events per predictor variable in the Cox model.
• When the primary outcome is a repeated continuous measure, use the expected within-individual standard deviation from published pilot data or a small internal pilot.

Handling Loss to Follow-Up

Classifying Attrition

Loss to follow-up is not a single phenomenon. Distinguish between:

• Administrative loss: owner relocation, practice transfer, study withdrawal
• Outcome-related loss: death, euthanasia, culling, or disease progression that prevents further measurement
• Measurement-specific loss: missed visits, failed assays, unusable imaging

Outcome-related loss is informative and must be modelled, not treated as random. Administrative loss is more likely to be ignorable, but this assumption requires verification. Compare baseline characteriztics of completers and non-completers. If they differ on prognostic variables, the missingness mechanism is not completely at random.

Mitigation Strategies

Schedule visits with buffer windows instead of fixed dates. A visit window of plus or minus two weeks accommodates owner schedules without introducing unacceptable measurement error for most outcomes. For production animals, align visits with existing husbandry procedures such as weighing, vaccination, or milking to reduce handling stress and labor costs.

Maintain a tracking log that records the reason for every missed visit. This log becomes the basis for sensitivity analyzes. When missing data exceed 20 percent, consider multiple imputation or inverse probability weighting, but report results under several missingness assumptions.

Documentation and Data Management

Protocol Documentation

The study protocol must specify the visit schedule, measurement procedures, and quality control steps before enrollment. The World Organization for Animal Health surveillance standards emphasize that data collection procedures must be standardized and documented to support international comparability. For studies that inform trade-related decisions, adherence to these standards is a prerequisite for regulatory acceptance.

Data Capture

Use electronic data capture with range checks and mandatory fields where feasible. Paper forms invite transcription errors and complicate audit trails. For field studies without reliable internet access, offline-capable applications with synchronisation functions are preferable to paper.

Record the following for every measurement:

• Animal identification and visit number
• Date and time of measurement
• Operator identifier
• Instrument identifier and calibration status
• Ambient conditions that could affect the measurement
• Any protocol deviations or adverse events

Quality Assurance

Schedule periodic audits of a random sample of records against source documents. Re-measure a subset of animals to quantify drift in operator technique. For imaging outcomes, have a second reader score a random 10 percent of studies and report inter-rater agreement. The awake animal fMRI literature highlights that technical factors such as motion, physiological noise, and anesthetic depth can confound longitudinal comparisons, so these parameters must be recorded at every session.

Recognized Complications and Failure Modes

Longitudinal studies in veterinary populations fail in predictable patterns. The most damaging is silent selection bias introduced at enrollment, when animals that remain available for repeated sampling differ systematically from those that do not. This is detected early by comparing baseline characteriztics of animals that complete the first follow-up visit against those that do not. A difference in age, breed, body condition, or disease severity at baseline signals that attrition is not random and that the planned analysis must account for it.

A second failure mode is measurement drift. Instruments calibrated once and used across months or years produce data that shift with operator fatigue, reagent lot changes, or equipment degradation. Detection requires periodic calibration checks and the inclusion of control samples or reference animals at regular intervals. When drift is confirmed, the corrective action is to model the time of measurement as a covariate or to standardize values against the control series.

A third failure mode is outcome misclassification that changes over time. Diagnostic criteria that are applied strictly at the first visit but more loosely later, or that are applied by different personnel without a shared protocol, generate spurious trends. Early detection depends on blinded re-reading of a random subset of records and on formal inter-observer agreement checks during the first weeks of data collection.

A fourth failure mode is the silent loss of the time origin. When the exact date of disease onset, exposure, or enrollment is not recorded with sufficient precision, time-to-event analyzes become unreliable. This is detected by auditing a sample of records for completeness of the date fields and by verifying that the time origin definition matches the research question.

Common Errors and Corrective Action

Less experienced investigators often treat every measurement occasion as an independent observation. This inflates sample size, underestimates standard errors, and produces false-positive findings. The corrective action is to specify the within-animal correlation structure in the analysis plan before data collection begins, using either a mixed-effects model or generalized estimating equations as described in the analysis section.

A second common error is over-sampling early and under-sampling late. Investigators schedule frequent visits when the outcome is unlikely to change and sparse visits when change is rapid. The corrective action is to map the expected trajectory of the outcome before finalising the visit schedule and to concentrate measurement effort at periods of expected change.

A third error is the collection of excessive data at each visit. Broad panels of tests generate missing values, increase cost, and distract from the primary outcome. The corrective action is to restrict each visit to the measurements that answer the primary question and to archive samples for future assays instead of running them immediately.

A fourth error is the failure to pre-specify the analysis of missing data. Investigators who decide how to handle missing values after seeing the results can unconsciously select the approach that favours their hypothesis. The corrective action is to state the primary analysis method for missing data in the protocol and to report sensitivity analyzes under alternative assumptions.

Limitations of the Current Evidence

The veterinary longitudinal literature is thinner than the human literature, and several design questions remain unresolved. The optimal frequency of measurement for slowly progressive conditions such as chronic kidney disease is not established, and published recommendations rest on expert opinion instead of comparative trials. The MSD Veterinary Manual provides species-specific guidance on monitoring intervals for common diseases, but the evidence base for these intervals is often observational.

Attrition patterns in veterinary populations are poorly characterized. Most published studies report loss to follow-up but few analyze the predictors of attrition or compare the characteriztics of completers and dropouts. The CDC principles of epidemiology describe general approaches to cohort maintenance, but these are oriented to human populations and may not transfer directly to animal populations where owners control access.

Expert opinion differs on the handling of animals that receive treatment during the follow-up period. Some investigators recommend censoring at the time of treatment, others recommend treating treatment as a time-varying covariate, and others recommend excluding treated animals entirely. Each approach answers a different question, and the choice should be made explicit in the protocol. The WOAH animal health surveillance standards and the WOAH terrestrial animal health code address surveillance design in production animal populations, but they do not resolve the treatment-censoring question for research studies.

Referral and Escalation

Most longitudinal studies do not require external consultation, but several circumstances warrant escalation. Statistical support should be sought before enrollment when the planned analysis involves complex correlation structures, competing risks, or joint modeling of repeated measures and survival. Laboratory involvement is required when assays are performed across multiple sites or when sample storage and batch effects could compromise comparability.

Regulatory reporting obligations arise when a study identifies a notifiable disease, a suspected adverse drug event, or a welfare concern that falls under local or national requirements. The AVMA practice resources provide guidance on professional obligations in the United States, while the WOAH terrestrial animal health code describes international notification standards for listed diseases. Investigators should identify the relevant reporting pathways before the study begins, not after a positive finding emerges.

Specialist consultation is warranted when the outcome measure requires equipment or expertise beyond the practice setting. Advanced imaging modalities such as functional magnetic resonance imaging in awake animals demand specialised facilities and personnel, and their use in longitudinal designs requires careful attention to repeated anesthesia, motion artefact, and habituation protocols as described in the methodological literature on mapping stress networks in awake animals. Similarly, longitudinal imaging of tumor vasculature requires serial contrast agent administration and standardized acquisition parameters, as demonstrated in longitudinal assessment of glioma vasculature using dynamic contrast MRI. When such resources are not available internally, referral to a collaborating institution is preferable to compromising the measurement protocol.

Frequently Asked Questions

How do I manage a longitudinal study when the ideal measurement instrument is unavailable or unaffordable?

Prioritize the research question over the instrument. If the reference standard is cost prohibitive, select a correlated surrogate that preserves the within-individual contrast you need. Validate the surrogate against the reference in a small pilot subset before full enrollment. Document the substitution explicitly in the protocol and in any resulting publications. Consider whether a commercial diagnostic laboratory can provide standardized assays, which may reduce equipment costs while shifting quality assurance responsibilities. For imaging or physiological monitoring, explore shared institutional equipment or collaborative arrangements with other research groups. The CDC principles of epidemiology emphasize that measurement validity and reliability matter more than technological sophistication, and a modest instrument used consistently across all visits outperforms a superior instrument used irregularly.

What are the realistic costs of a longitudinal study, and how should I budget for them?

Budget for three cost categories: fixed setup costs, per-visit variable costs, and hidden costs. Fixed costs include equipment, protocol development, and ethics approvals. Per-visit costs include consumables, staff time, animal handling, and data entry. Hidden costs include attrition replacement, instrument recalibration, data cleaning, and statistical consultation. A common error is underestimating staff time for follow-up coordination and reminder systems. Build a contingency of at least 15 to 20 percent of the total budget for unexpected losses or protocol amendments. For production animal populations, factor in the economic value of the animals and any productivity losses during sampling. The MSD Veterinary Manual provides species-specific guidance on handling and sampling procedures that can inform realistic time and cost estimates for each visit type.

How does longitudinal study design differ between companion animals and production animals?

Companion animal populations allow frequent, owner-mediated visits but suffer from owner-driven attrition and variable compliance with visit schedules. Production animal populations offer controlled environments and consistent management but impose constraints on handling frequency, withdrawal periods, and economic endpoints. In companion animals, the source population is often a single or small number of clinics, which limits generalizability. In production animals, the source population may be multiple herds or flocks, requiring cluster-based sampling and hierarchical analysis. Time scales also differ: companion animal studies may span years to capture age-related outcomes, while production animal studies often compress into weeks or months aligned with production cycles. The WOAH terrestrial animal health standards provide frameworks for surveillance in production populations that can inform sampling strategies and disease outcome definitions.

What should I do when a scheduled visit is missed or delayed?

Handle missed visits prospectively, not reactively. Define a visit window in the protocol, for example plus or minus 20 percent of the target interval, and classify any measurement outside that window as a protocol deviation. Record the actual date and time for every measurement, because time-varying covariates and outcomes require precise time stamps for analysis. If a visit is missed entirely, decide whether the animal remains eligible for subsequent visits or is censored from that point. For repeated measures analyzes, missing data mechanisms matter: if the probability of missingness depends on the outcome value, standard methods will be biased. The CDC epidemiology course distinguishes informative from non-informative missingness, and this distinction should drive your analytic approach. Consider whether a make-up visit within a defined window can recover the data point without violating the measurement schedule.

How do I explain the purpose and requirements of a longitudinal study to an owner or farm manager?

Frame the explanation around the value of repeated observation instead of the statistical machinery. Explain that a single measurement captures only a snapshot, while repeated measurements reveal how the animal changes over time and how individuals differ in their trajectories. Use a concrete example relevant to the audience, such as monitoring kidney function across the lifespan of a dog or tracking growth and feed efficiency across a production cycle. Be explicit about the time commitment, the number of visits, and what each visit involves. Describe what will happen if the animal becomes ill or is withdrawn. The AVMA practice resources offer guidance on client communication and informed consent that can be adapted for research contexts. Emphasize that the data contribute to better clinical decision-making for future patients, and clarify any financial arrangements, including whether the study covers examination or treatment costs.

How should I document protocol deviations and amendments during the study?

Maintain a deviation log separate from the case report forms. Record the date, animal identifier, nature of the deviation, reason, and corrective action taken. Review the log at regular intervals, for example monthly, to identify patterns that may indicate systemic problems such as unrealistic visit windows or inadequate staff training. Distinguish between minor deviations that do not affect data validity and major deviations that require statistical adjustment or exclusion. Amendments to the protocol itself require formal approval from the relevant ethics or institutional review body before implementation. The WOAH surveillance standards emphasize documentation as a core component of data quality, and this principle applies equally to research protocols. At study completion, summarize deviations and amendments in the final report so readers can assess the robustness of the findings.

Related Clinical & Scientific Guides

• Evaluating Veterinary Surveillance System Attributes
• Network Analysis for Infectious Disease Spread in Animal Populations
• Randomized Controlled Trials in Veterinary Field Settings

References and Further Reading

• Mapping stress networks using functional magnetic resonance imaging in awake animals.. 2018.
• Clinical crown length changes from age 12-19 years: a longitudinal study.. 2000.
• Neuroimaging research in human MDMA users: a review.. 2007.
• College-Aged Males Experience Attenuated Sweet and Salty Taste with Modest Weight Gain.. 2017.
• Chronic kidney disease in dogs in UK veterinary practices: prevalence, risk factors, and survival.. 2013.
• High-resolution longitudinal assessment of flow and permeability in mouse glioma vasculature: Sequential small molecule and SPIO dynamic contrast agent MRI.. 2009.
• WOAH Animal Health Surveillance Standards. WOAH.
• CDC Principles of Epidemiology in Public Health Practice. CDC.
• MSD Veterinary Manual, Professional Edition. MSD Veterinary Manual.

Related Articles

• Designing Cross-Sectional Studies in Veterinary Populations
• Cohort Studies in Veterinary Medicine: Design and Analysis
• Understanding Ecological Studies in Veterinary Epidemiology
• Case-Control Studies in Veterinary Epidemiology: Selection and Analysis
• Cluster Sampling in Veterinary Field Studies

This article is educational professional reference material for veterinary audiences. It is not a substitute for veterinary diagnosis, individual clinical judgment, current product labeling, or applicable regulatory requirements.