Cross-Over Designs in Veterinary Clinical Trials: Advantages and Limitations

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

Key Takeaways

• Cross-over designs offer enhanced statistical efficiency by utilizing each animal as its own control, thereby eliminating between-subject variability. This is particularly advantageous for chronic, stable conditions with low within-animal variability, potentially reducing sample size requirements compared to parallel-group designs.
• The primary threat to validity is carryover, the persistent effect of the first treatment into the second period, which can be pharmacological (e.g., drug metabolites) or physiological (e.g., receptor upregulation). Adequate washout periods, informed by pharmacokinetic data (e.g., elimination half-life) and pharmacodynamic considerations, are crucial to mitigate this.
• Period effects, systematic differences between trial periods unrelated to treatment (e.g., disease progression, seasonal changes), can confound results. While randomizing treatment order balances period effects across sequences, interaction between period effects and treatment is indistinguishable from carryover in two-period designs.
• Cross-over designs are best suited for outcomes with high within-animal repeatability and minimal spontaneous change, such as objective measurements like serum biochemistry or blood pressure. They are poorly suited for acute, self-limiting, or progressive diseases, or outcomes with significant learning or irreversible changes, such as some behavioral assessments.
• Robust reporting, adhering to guidelines like ARRIVE, is essential, detailing the rationale for design choice, washout justification (pharmacokinetic/pharmacodynamic), and methods for assessing carryover and period effects. Statistical analysis should employ mixed-effects models accounting for the paired nature of data and within-animal correlation.
• Failure modes such as incomplete washout, differential dropout between sequences, and protocol violations can be detected through proactive monitoring. For instance, incomplete washout can be identified by measuring analyte levels at the end of the washout period and comparing them to baseline.

This article examines the cross-over design as applied to veterinary clinical research. It serves veterinary researchers, graduate students, and clinicians who appraise trial reports or design their own studies. The focus is on the structural logic of the design, the statistical and biological assumptions it requires, and the specific failure modes that threaten validity in animal populations. Parallel-group designs are addressed only as contrast points where they clarify a cross-over property.

The central question this article answers is practical: when does a cross-over design offer a genuine advantage over a parallel-group design in veterinary research, and what conditions must hold for its results to be credible? The answer depends on understanding carryover, period effects, and the biological characteriztics of the species and outcome under study. These concepts determine whether a cross-over trial will produce estimates that generalize to the target population or whether it will produce artefactual differences that mislead.

At a Glance

Conceptual Foundations of the Cross-Over Design

A cross-over trial administers each intervention to the same experimental unit in a sequence, with the order randomised. The simplest form is the AB/BA design, in which animals receive treatment A then treatment B, or the reverse order. Each animal therefore contributes a paired comparison, and the treatment effect is estimated from within-subject differences instead of between-group differences.

The statistical appeal is immediate. Between-subject variability, often the dominant source of noise in veterinary outcome data, is removed from the treatment contrast. Body weight, breed, age, baseline disease severity, and individual metabolic variation no longer confound the comparison because each animal is compared with itself. This property can reduce the required sample size substantially when the outcome is highly variable between animals but relatively stable within an animal over time.

The design logic depends on a set of assumptions that must be examined before the trial begins. The first is that the effect of a treatment administered in the first period does not persist into the second period. The second is that the condition under study is stable across the trial duration. The third is that the outcome measurement is repeatable and does not change systematically with repeated testing. When these assumptions hold, the cross-over design is statistically efficient and ethically attractive because it requires fewer animals than a parallel-group trial.

Carryover Effects

Carryover is the persistence of a treatment effect into a subsequent period. It is the most serious threat to the validity of a cross-over trial because it biases the estimate of the second-period treatment effect in a way that cannot be fully corrected by statistical adjustment. The direction and magnitude of the bias depend on the nature of the residual effect and the sequence in which treatments were administered.

Pharmacological carryover occurs when the drug or its active metabolites remain in the animal beyond the intended washout. This is a matter of pharmacokinetics and can usually be managed with an adequate washout interval derived from elimination half-life data. More subtle is physiological carryover, in which the drug has produced a lasting change in the animal's biology. Examples include receptor upregulation or downregulation, enzyme induction, immune modulation, or structural tissue change. A drug that alters renal function, for instance, may change the clearance of a subsequently administered agent even after the first drug is no longer detectable.

Disease-related carryover is a distinct category. If the first treatment cures or substantially improves the condition, the animal entering the second period is no longer comparable to its first-period state. This is particularly problematic in infectious disease trials where sterilizing cure is possible, or in surgical trials where the intervention produces permanent anatomical change. In such cases the cross-over design is simply inappropriate, and a parallel-group design must be used.

The statistical consequence of carryover is that the treatment effect estimate becomes a weighted average of the true treatment difference and the carryover difference, with the weights determined by the sequence groups. The standard test for carryover, based on comparing the sums of the two period measurements between sequence groups, has low power in the typical sample sizes used in veterinary trials. A non-significant carryover test does not establish the absence of carryover, it may simply reflect insufficient power to detect it.

Period Effects

A period effect is a systematic difference between the first and second trial periods that affects all animals equally, regardless of treatment. Sources include disease progression, seasonal variation, growth and development, environmental changes, and observer drift in outcome assessment. In production animals, weight gain over the trial period can obscure treatment effects on growth-related outcomes. In companion animals, behavioral outcomes may change as animals habituate to the study environment or to the measurement procedures.

The AB/BA design confounds treatment with period in a specific way. The treatment effect is estimated from the difference between periods, so any systematic period effect is subtracted out when the two sequence groups are balanced. The design remains valid for estimating the treatment difference even in the presence of period effects, provided the period effect is additive and identical across sequence groups. The problem arises when the period effect interacts with treatment, meaning the response to treatment B differs depending on whether it was given first or second. This interaction is statistically indistinguishable from carryover in the two-period design.

Period effects are best managed by design instead of analysis. Keeping the trial duration short reduces the opportunity for temporal drift. Including baseline measurements at the start of each period allows period-specific change to be estimated. Randomising the order of treatments ensures that period effects are balanced across sequence groups, but it does not remove them. The researcher must judge whether the condition under study is sufficiently stable over the planned trial duration for the design to be credible.

Within-Subject Stability and Outcome Repeatability

The efficiency of a cross-over design depends on the correlation between repeated measurements on the same animal. When within-animal variability is low relative to between-animal variability, the design delivers its promised reduction in sample size. When the outcome is noisy, with large day-to-day variation within an individual, the advantage shrinks and may disappear entirely.

Objective outcomes such as serum biochemistry, blood pressure, or imaging parameters tend to be more repeatable than behavioral or subjective clinical scores. Pain assessment in particular poses challenges. Facial expression scoring, as described in the equine pain face literature, can detect pain-related changes, but the expression of pain may vary with the presence of observers and other environmental factors. A cross-over trial using such outcomes must demonstrate that the measurement tool is repeatable across the trial periods and that the animals' baseline state is stable.

Learning effects are a specific form of period effect relevant to behavioral and cognitive outcomes. An animal tested on a task in the first period may perform better in the second period simply because it has learned the task. This systematically inflates second-period scores regardless of treatment and can mask or exaggerate treatment differences. The problem is well recognized in human cognitive trials, where practice effects on memory tasks are substantial. Veterinary researchers studying behavior or cognition must either use outcomes with demonstrated test-retest reliability or incorporate design features that account for learning.

Reporting Standards and Design Justification

Veterinary cross-over trials should be reported with sufficient detail for readers to assess the validity of the design and the credibility of the conclusions. The ARRIVE guidelines specify the minimum information required for transparent reporting of animal research, including allocation concealment, blinding, sample size calculation, and statistical methods. The EQUATOR Network maintains a comprehensive library of reporting guidelines that includes resources relevant to randomised trials and observational studies. Researchers should consult these standards at the design stage, not after data collection is complete.

The report should state the rationale for choosing a cross-over design, the expected within-subject correlation, the planned washout duration and its pharmacokinetic justification, and the methods used to assess carryover and period effects. Authors should report both period-specific results and the overall treatment estimate, and they should discuss the plausibility of carryover in the context of the drug's pharmacology and the disease's natural history. A cross-over trial that does not address these elements cannot be interpreted with confidence, regardless of the statistical significance of its primary outcome.

Washout Periods and Residual Treatment Effects

The washout interval between treatment periods serves a single purpose: it allows the physiological effect of the first intervention to dissipate before the second period begins. A washout that is too short permits residual pharmacological activity to contaminate the second period, producing a carryover effect that is indistinguishable from a true treatment effect in the analysis. A washout that is unnecessarily long increases the risk that the patient's underlying condition changes between periods, which inflates the period effect and reduces the precision advantage of the cross-over design.

The correct washout duration depends on the pharmacokinetics of the intervention, not on convention. For systemically absorbed drugs, the washout should be derived from the elimination half-life of the active moiety and any active metabolites. A common working rule is five half-lives for the parent compound, but this assumes linear kinetics and ignores tissue sequestration. For drugs with long terminal elimination phases, such as lipophilic compounds that accumulate in adipose tissue, five half-lives may be insufficient. For locally acting products, such as topical dermatological preparations or intra-articular injections, the relevant parameter is the residence time at the site of action, which may be considerably longer than the plasma half-life.

The investigator should also consider whether the intervention produces a biological effect that outlasts drug exposure. An anti-inflammatory that modifies gene expression, an immunomodulatory agent that alters cell populations, or a dietary intervention that changes the gut microbiome may produce effects that persist for weeks after the compound is cleared. In such cases, pharmacokinetic half-life is a poor guide, and the washout must be justified on pharmacodynamic grounds. The ARRIVE guidelines require that the rationale for washout duration be reported explicitly, and reviewers should treat an unreasoned washout as a design flaw.

When a washout is omitted entirely, the design becomes a two-period cross-over with no interval between treatments. This is only defensible when the outcome is measured during the intervention and the effect is known to reverse rapidly upon cessation. The SWAP-MEAT trial, which compared plant-based and animal-based meat products over two eight-week periods with no washout, illustrates this approach in human nutrition research. The investigators reasoned that the primary outcome, fasting serum trimethylamine-N-oxide, would respond to dietary change within days, making a washout unnecessary. In veterinary research, a no-washout design might be considered for dietary interventions or behavioral studies where the outcome tracks the intervention closely, but the burden of proof rests on the investigator to demonstrate that residual effects are negligible.

Sequence Allocation and Blinding

The order in which treatments are administered must be randomised, not assigned by convenience. Randomisation of sequence protects against allocation bias and provides the probabilistic foundation for the statistical analysis. In a simple AB/BA design, each animal is randomly assigned to receive treatment A then B, or treatment B then A. For three-period designs, all six possible sequences should be considered, and for four-period designs the number of permutations grows rapidly. The randomisation list should be generated by a computer algorithm, not by coin toss or alternation, and the allocation schedule should be concealed from the enrolling clinician.

Blinding in a cross-over trial operates at two levels. The investigator who administers the treatment and assesses the outcome should be masked to the treatment identity, and the animal's caretakers should be masked where feasible. Blinding is particularly important in cross-over designs because the same animal serves as its own control, and the assessor's expectation of a treatment effect can bias subjective outcome measures. The equine pain face study by Gleerup and colleagues used a semi-randomised cross-over design in which pain was induced by tourniquet and capsaicin application, and facial expressions were scored from video recordings. The use of video scoring allowed the assessor to be blinded to the treatment condition, which strengthened the validity of the pain face descriptions. In veterinary trials, video recording, automated data capture, and laboratory assays on coded samples all facilitate blinding when the intervention itself cannot be masked.

Blinding is impossible when the treatments are physically dissimilar, such as a surgical procedure compared with medical management, or when the route of administration differs. In these situations, the investigator should consider a blinded outcome assessor who is not involved in treatment administration. The EQUATOR Network maintains reporting guidelines that specify how blinding should be described, and the CONSORT extension for crossover trials requires explicit reporting of who was blinded and how blinding was achieved.

Statistical Analysis of Cross-Over Data

The analysis of a two-period, two-treatment cross-over trial must account for the paired nature of the data and the potential for period effects. The standard approach is a mixed-effects model with fixed effects for treatment, period, and sequence, and a random effect for animal. This model partitions the total variance into between-animal and within-animal components, and the treatment effect is estimated from the within-animal contrast. The period effect captures any systematic difference between the first and second observation, such as disease progression, seasonal variation, or learning effects in behavioral outcomes. The sequence effect, when tested, provides a check for carryover, although the test has low power and a non-significant result does not prove the absence of carryover.

The analysis should be planned before the trial begins, and the primary outcome and its analysis model should be specified in the protocol. Post hoc decisions about whether to include or exclude period effects, or whether to analyze the first period alone, undermine the validity of the conclusions. If carryover is suspected, the recommended approach is to analyze only the first-period data, which is equivalent to a parallel-group comparison, but this sacrifices the precision advantage of the cross-over design and should be presented as a sensitivity analysis instead of the primary result.

Sample size calculation for a cross-over trial requires an estimate of the within-animal standard deviation of the outcome, not the between-animal standard deviation. The within-animal variability is typically smaller, which is why cross-over designs require fewer animals than parallel-group designs for the same power. However, the within-animal standard deviation is often unknown at the planning stage, and the investigator should use published values from similar studies or conduct a pilot study. The MSD Veterinary Manual and other species-specific references may provide baseline values for common clinical outcomes, but the investigator should be cautious about extrapolating variability estimates across breeds, ages, and disease states.

Decision Framework for Cross-Over Versus Parallel-Group Design

The decision to use a cross-over design rests on three conditions. First, the condition under study must be chronic and stable, so that the animal's baseline status does not change materially across the trial duration. Second, the intervention must have an effect that is reversible and does not permanently alter the disease process. Third, the outcome must be measurable with acceptable within-animal repeatability. When these conditions are met, the cross-over design offers a substantial reduction in sample size and an increase in statistical power. When they are not met, the parallel-group design is the safer choice.

The following table summarizes the decision criteria.

A worked example illustrates the decision process. Consider a trial of a new non-steroidal anti-inflammatory drug for canine osteoarthritis. The disease is chronic and stable over a three-month horizon, the drug's effect is expected to reverse within days of discontinuation, and the pharmacokinetic half-life supports a two-week washout. Gait analysis provides a continuous outcome with acceptable repeatability. The cross-over design is appropriate, and a sample size of 12 dogs may provide adequate power. Contrast this with a trial of a disease-modifying therapy for feline chronic kidney disease. The treatment is intended to slow disease progression, the effect is not expected to reverse upon discontinuation, and the disease itself progresses over months. A cross-over design would be inappropriate because the first-period treatment would alter the trajectory of the disease, and the second-period observations would be confounded by disease progression. A parallel-group design is required.

Species and production system considerations also influence the decision. In food animals, withdrawal periods for drugs used in the first period may extend the washout beyond a practical duration, and the cost of holding animals through multiple periods may be prohibitive. In wildlife research, capture and handling for repeated sampling may induce stress responses that confound the outcome, and the risk of loss to follow-up may be high. In companion animals, client compliance with repeated visits and medication administration is a practical constraint. The AVMA practice resources provide guidance on client communication and compliance that is relevant to the planning of multi-period trials. The WOAH terrestrial animal health standards address welfare and ethical considerations that apply to repeated interventions in research animals, and these standards should be consulted when the trial involves multiple procedures per animal.

Recognized Failure Modes and Early Detection

Cross-over trials in veterinary medicine fail most often through mechanisms that are detectable before data analysis, provided the trial is monitored with explicit stopping rules. The most common failure modes are incomplete washout, differential dropout between sequences, and protocol violations that cluster in one period.

Incomplete washout is detected by measuring the analyte or outcome of interest in a subset of animals at the end of the washout period and comparing values with baseline. If a proportion of animals exceed a pre-specified threshold, for example 20% of the baseline value for a pharmacokinetic endpoint, the washout is inadequate and the trial should be paused. For behavioral or pain outcomes, where no measurable residue exists, the investigator must rely on published pharmacodynamic duration data and should extend the washout by at least two half-lives of the pharmacodynamic effect, not the parent drug.

Differential dropout is identified by tracking the reasons for withdrawal by sequence group. If animals withdraw more often during the second period of one sequence than the other, the cause is likely to be period-related, such as disease progression or seasonal change, instead of treatment-related. The analysis should then include a per-protocol comparison alongside the intention-to-treat analysis, and the discrepancy between them should be reported explicitly.

Protocol violations that cluster in one period are detected by reviewing compliance logs before unblinding. A violation rate above 10% in a single period, or a pattern where one technician administered all treatments in that period, should trigger a review of the standard operating procedures before the data are analyzed.

Common Errors in Design and Execution

Less experienced investigators frequently underestimate the importance of the no-treatment or placebo period. A cross-over design without a control period cannot distinguish a treatment effect from a period effect, and the analysis will be uninterpretable if the condition under study fluctuates spontaneously. The corrective action is to include a placebo or sham-treatment period in every sequence, even when the active treatment is expected to have a large effect.

A second recurring error is the use of a cross-over design for an outcome with a strong learning or habituation component. Pain scoring, gait analysis, and behavioral assessments all improve with repeated exposure to the testing environment. The corrective action is to run a familiarisation session before the first treatment period and to randomise the order of treatments within each sequence, also the sequence allocation itself.

A third error is the analysis of cross-over data as if the observations were independent. Ignoring the within-subject correlation inflates the type I error rate. The corrective action is to specify the mixed-effects model in the statistical analysis plan before enrollment, including the random intercept for animal and the fixed effects for period, sequence, and treatment.

Limitations of the Current Evidence

The veterinary literature contains few cross-over trials with adequate reporting of carryover assessment and period effects. Many published trials in companion animals use sample sizes below ten, and the statistical power to detect a carryover effect is correspondingly low. The ARRIVE guidelines require reporting of the experimental unit, the randomisation procedure, and the blinding methods, but they do not mandate a formal test for carryover, and many authors omit this analysis entirely.

Expert opinion differs on the acceptable length of washout for behavioral outcomes. Some investigators argue that a washout of two weeks is sufficient for most analgesic and behavioral interventions, while others recommend a period equal to at least five times the duration of the observed effect. The EQUATOR Network maintains reporting checklists that can help authors identify missing elements, but these checklists do not resolve the underlying biological uncertainty about the duration of pharmacodynamic effects in different species.

There is also genuine disagreement about whether the AB/BA design should ever be used in veterinary medicine. The design is efficient but provides no unbiased estimate of carryover when carryover is present. For conditions with a variable natural history, such as canine atopic dermatitis or feline interstitial cystitis, the MSD Veterinary Manual advises that parallel-group designs are often more robust, and this view is widely held among veterinary dermatologists and behaviorists.

Referral, Consultation, and Regulatory Reporting

A veterinary researcher should seek statistical consultation before finalising the protocol, not after data collection. A biostatistician with experience in repeated-measures designs should review the planned sample size, the randomisation scheme, and the analysis model. If the trial involves a regulated product, the investigator should consult the relevant national authority early in the planning phase, because the requirements for demonstrating bioequivalence or residual efficacy differ from those for a proof-of-concept study.

Laboratory involvement is warranted when the outcome depends on assay performance. Hormone assays, cytokine panels, and metabolomic profiles are subject to batch effects, and the laboratory should be asked to run all samples from both periods in a single batch or to provide a plan for batch correction.

Regulatory reporting is required when an adverse event occurs during a treatment period and the investigator cannot exclude a causal relationship with the investigational product. The WOAH terrestrial animal health standards and the AVMA practice resources both describe the obligations of investigators to report serious adverse events promptly, and the investigator should have a written adverse event plan in place before the first animal is enrolled.

Frequently Asked Questions

How do I justify a cross-over design to a funding body or ethics committee when the condition is chronic and stable?

State the statistical efficiency gain explicitly. A cross-over design uses each animal as its own control, which removes between-subject variance from the treatment comparison. For chronic, stable conditions with low within-subject variability, this reduces the required sample size substantially compared with a parallel-group design. Reference the ARRIVE guidelines when preparing the justification, as they require clear description of study design and sample size rationale. Acknowledge the washout period and the risk of carryover in your protocol. Ethics committees will expect evidence that the washout duration is justified by the drug's pharmacokinetic profile and that sequence allocation is randomised.

What are the minimum washout parameters I should specify for a novel compound with unknown pharmacokinetics?

You cannot assume a standard washout. The minimum is five times the terminal elimination half-life of the parent compound and any active metabolites, measured in the target species. If the half-life is unknown, you must run a pilot pharmacokinetic study first. For drugs with enterohepatic recirculation or tissue accumulation, longer washouts are needed. The MSD Veterinary Manual provides species-specific pharmacological reference values that can inform these decisions. When the half-life cannot be determined, consider a parallel-group design instead. Document your reasoning in the protocol, including the assay method and the limit of quantification used to confirm that baseline concentrations have returned to pre-treatment levels.

How should I handle missing data when an animal is withdrawn during the second period?

The analysis strategy must be specified before the trial begins. If the withdrawal is unrelated to treatment, a complete-case analysis restricted to animals finishing both periods is acceptable but reduces power. If the withdrawal is treatment-related, complete-case analysis introduces bias. Use a mixed-effects model that accommodates missing data under a missing-at-random assumption, and report sensitivity analyzes under different missingness assumptions. State the withdrawal criteria in the protocol, including objective thresholds for adverse events. The EQUATOR Network hosts reporting guidelines that specify how participant flow and missing data should be described in the final manuscript. Do not impute outcomes without a prespecified imputation model.

Can I use a cross-over design in a food-producing species with withdrawal period constraints?

Yes, but the washout period must satisfy both pharmacological and regulatory requirements. The washout must be at least as long as the labelled withdrawal period for the longest-acting treatment, and it must also be long enough to eliminate carryover. These two requirements may conflict. When the regulatory withdrawal period exceeds the pharmacological washout, the design becomes logistically difficult and expensive. Consult the WOAH terrestrial animal health standards for international expectations regarding residue testing and animal welfare during the trial. In production animals, the cost of housing animals through extended washouts often makes a parallel-group design more practical. Document the regulatory basis for your washout decision in the study protocol.

What records should I keep to demonstrate that the washout was effective?

Maintain individual animal logs that include the date and time of each treatment administration, the date of the final sample collection for the first period, and the date of the first sample collection for the second period. Record any intercurrent illness and all concomitant medications, since these can prolong or shorten effective washout. Store the pharmacokinetic data that justify the washout duration, including the assay validation report and the limit of quantification. For each animal, document that pre-treatment baseline values in the second period are within the expected range for that individual. The AVMA practice resources provide guidance on medical record standards that apply to clinical research records as well as clinical care.

How do I explain the limitations of a cross-over design to a client whose animal is enrolled?

Explain that the animal will receive both treatments in sequence, with a rest period between them. The rest period exists to ensure the first treatment has cleared the body before the second begins. The main risk is that the first treatment could still be influencing the animal when the second period starts, which would make the results difficult to interpret. Emphasize that the design reduces the number of animals needed, because each animal serves as its own comparison. Be transparent that the total study duration is longer for each participant than in a parallel-group trial. The ARRIVE guidelines emphasize that animal welfare considerations should be reported, and you can reassure owners that their animal's welfare monitoring is part of the study protocol.

Related Clinical & Scientific Guides

• Conducting Systematic Reviews of Veterinary Diagnostic Test Accuracy
• Bias in Veterinary Research: Types, Sources, and Mitigation
• Cluster Randomized Trials in Veterinary Research: Design and Analysis

References and Further Reading

• An equine pain face.. 2015.
• Piragliatin (RO4389620), a novel glucokinase activator, lowers plasma glucose both in the postabsorptive state and after a glucose challenge in patients with type 2 diabetes mellitus: a mechanistic study.. 2010.
• Effect of allopurinol on blood pressure of adolescents with newly diagnosed essential hypertension: a randomized trial.. 2008.
• Effect of estrogen on brain activation patterns in postmenopausal women during working memory tasks.. 1999.
• Citrus polyphenol hesperidin stimulates production of nitric oxide in endothelial cells while improving endothelial function and reducing inflammatory markers in patients with metabolic syndrome.. 2011.
• A randomized crossover trial on the effect of plant-based compared with animal-based meat on trimethylamine-N-oxide and cardiovascular disease risk factors in generally healthy adults: Study With Appetizing Plantfood-Meat Eating Alternative Trial (SWAP-MEAT).. 2020.
• ARRIVE Guidelines 2.0 for Reporting Animal Research. PLOS Biology, 2020.
• EQUATOR Network Reporting Guidelines. EQUATOR Network.
• MSD Veterinary Manual, Professional Edition. MSD Veterinary Manual.

Related Articles

• Designing Crossover Trials for Veterinary Therapeutics
• Implementing Adaptive Trial Designs in Veterinary Clinical Research
• Pragmatic vs Explanatory Trials in Veterinary Clinical Research
• Designing Adaptive Clinical Trials for Veterinary Medicine
• Implementing Bayesian Methods in Veterinary Clinical Trials

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.