# Designing Dose-Response Studies in Veterinary Pharmacology ## Key Takeaways - Dose-response studies are fundamental for establishing the relationship between drug dose and biological effect, guiding rational drug selection in veterinary medicine by defining efficacy and safety profiles. - The dose range must bracket the anticipated therapeutic window, extending from below the minimum effective dose to above the maximum tolerated dose, informed by preliminary potency estimates or pilot studies. - Logarithmic dose spacing is preferred to efficiently characterize the steep portion of the sigmoidal dose-response curve, ensuring more informative data points in the clinically relevant range. - Objective endpoints (e.g., serum drug concentration, tumor volume) offer greater statistical power than subjective endpoints (e.g., pain scores), and surrogate endpoints require species-specific validation against clinical outcomes. - Blinding (double-blinding where feasible) and the inclusion of appropriate control groups (placebo, active comparator) are critical for preventing bias and establishing causal inference in dose-response studies. - Sample size calculations must account for expected effect size, endpoint variability, and desired statistical power, with regression-based analyses generally preferred over pairwise comparisons for curve estimation. --- Dose-response studies define the relationship between administered dose and the magnitude of a biological effect, forming the empirical basis for rational drug selection in veterinary medicine. This article provides a procedural framework for designing such studies across species, from initial dose selection through statistical analysis and reporting. It serves veterinary researchers planning preclinical efficacy trials, clinicians designing institutional investigations, and graduate students preparing study protocols. The content addresses the decisions that determine whether a dose-response study will yield interpretable, publishable data, including dose spacing, endpoint selection, sample size calculation, and common design failures. A well-constructed dose-response study answers a specific question: what dose produces the desired effect with acceptable toxicity in the target species? The answer depends on the shape of the dose-response curve, the therapeutic index of the compound, and the variability of the response within the population. These factors differ between companion animals, food animals, and laboratory species, and the design must accommodate those differences. Regulatory expectations for dose confirmation also vary by jurisdiction and by whether the drug is intended for food-producing animals, where tissue residue and withdrawal considerations impose additional constraints on study design. The principles described here apply to all veterinary species, but the emphasis on specific design elements will differ. Studies in food animals must consider withdrawal periods and extra-label use restrictions, while studies in companion animals may prioritize client-reported outcomes alongside objective measures. Laboratory animal studies often allow more invasive sampling and terminal endpoints. The researcher should consult current regulatory guidance from the relevant authority before finalizing a protocol, as requirements evolve and differ between regions. The [ARRIVE guidelines for reporting animal research](https://arriveguidelines.org/) provide a baseline for transparency that applies across species and study types. ## At a Glance | Parameter | Decision Required | Key Consideration | |---|---|---| | Study objective | Efficacy, safety, or both | Determines endpoint selection and dose range | | Dose range | Lowest to highest dose | Must bracket the anticipated therapeutic window | | Number of dose levels | 3 to 6 typically | More levels improve curve estimation, increase cost | | Dose spacing | Arithmetic or logarithmic | Logarithmic spacing detects steep curves efficiently | | Endpoint selection | Continuous or categorical | Continuous endpoints offer greater statistical power | | Control group | Placebo, active comparator, or both | Required for causal inference | | Sample size | Per dose group | Based on expected effect size and variability | | Species and model | Target species or surrogate | Target species preferred for clinical relevance | | Statistical analysis | Regression or ANOVA | Pre-specified in the protocol before data collection | ## The Pharmacological Basis of Dose-Response Design The dose-response relationship reflects the interaction between a drug and its biological target, governed by receptor occupancy, signal transduction, and homeostatic counter-regulation. For most drugs, the relationship between dose and effect follows a sigmoidal curve when plotted on a logarithmic dose axis. The steep portion of the curve corresponds to the range where small dose changes produce large effect changes, and it is within this region that dose selection has the greatest clinical impact. Below the threshold dose, no measurable effect occurs, above the plateau dose, increasing the dose produces no additional benefit but may increase toxicity. The therapeutic index, the ratio between the dose producing toxicity and the dose producing the desired effect, determines how forgiving the dose-response relationship is in clinical use. A drug with a narrow therapeutic index requires more precise dose selection and justifies a study design with closely spaced dose levels. A drug with a wide therapeutic index may be adequately characterized with fewer, more widely spaced doses. The researcher should estimate the expected therapeutic index from preclinical data, published literature, or structurally related compounds before selecting the dose range. Pharmacokinetic considerations interact with the dose-response relationship in ways that affect study design. A drug with nonlinear pharmacokinetics may show a disproportionate increase in plasma concentration with increasing dose, steepening the observed dose-response curve. A drug with a long half-life may require a loading dose or a longer observation period before steady-state effects can be measured. The design must account for the time course of drug distribution and elimination relative to the timing of endpoint measurement. ## Dose Selection and Spacing The dose range must bracket the anticipated therapeutic window, extending below the expected minimum effective dose and above the expected maximum tolerated dose. This requires a preliminary estimate of potency, which can be derived from in vitro receptor binding data, pharmacokinetic modeling, or published studies in related species. When no prior data exist, a pilot study with a small number of animals per dose level can refine the range before the full study begins. The number of dose levels represents a trade-off between curve resolution and resource allocation. Three dose levels can establish that a relationship exists but cannot reliably estimate the shape of the curve. Four to six dose levels allow fitting of a full sigmoidal model and estimation of the half-maximal effective dose with acceptable precision. The [ELLDOPA trial design](https://pubmed.ncbi.nlm.nih.gov/16222436/), a dosage-ranging study in human Parkinson's disease, used four treatment groups including placebo and three active dose levels, a structure that translates directly to veterinary applications. Dose spacing should follow a logarithmic scale for most drugs, with each successive dose being a constant multiple of the previous one. Common spacing factors are 2, 3, or 10, depending on the expected steepness of the curve. Logarithmic spacing places more dose levels in the steep, informative portion of the curve and prevents the common error of clustering doses in the plateau region. Arithmetic spacing is appropriate only when the dose-response curve is known to be shallow across the range of interest. ## Endpoint Selection and Measurement The primary endpoint must be biologically meaningful, measurable with acceptable precision, and sensitive to dose-related changes. Objective endpoints such as serum drug concentration, tumor volume, or bacterial count offer greater statistical power than subjective endpoints such as pain scores or activity levels. When subjective endpoints are unavoidable, they should be measured with validated instruments, by observers blinded to treatment assignment, and at standardized time points. Surrogate endpoints may shorten study duration or reduce animal numbers but require validation against clinically relevant outcomes. A biomarker that correlates with disease progression in one species may not correlate in another, and the researcher should justify the choice of surrogate endpoint with species-specific evidence. The relationship between a surrogate and a clinical outcome can itself be characterized in a dose-response study, as illustrated by investigations of [growth factors for intervertebral disc degeneration](https://pubmed.ncbi.nlm.nih.gov/15564925/), where biochemical and histological outcomes served as surrogates for functional recovery. The timing of endpoint measurement must align with the time course of drug action. A single measurement at a fixed time point may miss the peak effect or capture it inconsistently across dose groups if the time to peak effect differs by dose. Repeated measurements at multiple time points provide a more complete picture but increase the statistical complexity of the analysis and the burden on the animals. The protocol should specify the primary analysis time point and treat other measurements as secondary outcomes. ## Control Groups and Blinding A placebo control group is necessary to distinguish drug effects from spontaneous recovery, regression to the mean, and observer bias. The placebo group also provides the baseline against which the lowest dose is compared, allowing detection of a minimum effective dose. An active comparator group, using an approved drug for the same indication, can provide a clinical benchmark and help position the new drug relative to existing therapy. Blinding prevents conscious or unconscious bias from influencing endpoint measurement, particularly for subjective outcomes. Double-blinding, where neither the investigator nor the animal caretaker knows the treatment assignment, is the standard for clinical trials. Blinding is feasible even in surgical studies through the use of sham procedures, and it should be maintained until the database is locked and the primary analysis is complete. The [reporting standards for animal research](https://arriveguidelines.org/) require explicit description of blinding procedures, and reviewers increasingly expect this information. ## Sample Size and Statistical Power Sample size calculation requires an estimate of the expected effect size, the variability of the endpoint, and the desired statistical power. The effect size should be the smallest difference that would be clinically meaningful, not the largest difference the investigator hopes to detect. Variability estimates can be obtained from pilot data, published studies in the same species and model, or conservative assumptions based on the endpoint type. The statistical analysis plan should be specified before data collection begins. For dose-response studies, regression-based approaches that model the relationship between log dose and effect are generally preferred over pairwise comparisons between individual dose groups. Regression analysis uses all the data to estimate the dose-response curve and provides estimates of the half-maximal effective dose and its confidence interval. Pairwise comparisons may supplement the regression analysis but should not replace it as the primary analytical approach. The number of animals per dose group depends on the variability of the endpoint and the steepness of the dose-response curve. Studies with highly variable endpoints or shallow curves require larger group sizes to detect a given difference. The researcher should also consider the potential for attrition, particularly in studies with long observation periods or invasive procedures, and inflate the sample size accordingly. ## Dose Titration and Adaptive Design Strategies Fixed-dose parallel designs remain the most common approach in veterinary dose-response work, but they are not always the most efficient. Dose titration designs, in which each subject receives sequentially adjusted doses until a predefined endpoint is reached, reduce the number of animals required and are particularly suited to chronic conditions where the therapeutic index is wide. The ELLDOPA trial in human Parkinson's disease illustrates the value of a parallel dosage-ranging design with gradual escalation, in that it permitted both efficacy and toxicity to be assessed within a single protocol while maintaining blinding across four treatment arms. The same logic transfers to veterinary trials in degenerative or progressive diseases, where a washout period between dose levels may be impractical. Adaptive designs, including Bayesian dose-escalation schemes, allow the dose assigned to the next subject to depend on the responses of previously enrolled subjects. These designs are attractive in veterinary oncology and in rare or severe conditions where patient numbers are limited. They require pre-specified decision rules, a statistician familiar with the chosen framework, and careful attention to the fact that early stopping rules based on toxicity may leave the efficacy estimate imprecise. A pragmatic middle ground is the two-stage design: a small pilot phase to establish tolerability and refine dose spacing, followed by a confirmatory phase with a fixed set of doses. The pilot phase should be reported separately and its results should not be pooled with the confirmatory data without a pre-specified analysis plan. Cross-over designs deserve consideration when the drug has a short half-life, the condition is stable, and carryover effects can be controlled by washout periods. They are statistically efficient because each animal serves as its own control, but they are vulnerable to period effects and to disease progression over the study interval. In food-producing animals, withdrawal periods may make cross-over designs impractical because the washout interval becomes prohibitively long. In companion animals with chronic stable disease, however, a cross-over with three or four dose levels can provide a complete dose-response curve from a cohort of 12 to 20 animals, whereas a parallel design would require three to four times that number. ## Dose-Response Analysis and Model Selection The analysis of dose-response data should begin with a graphical exploration of the raw responses plotted against dose on a logarithmic scale. This step reveals the shape of the relationship, the presence of outliers, and whether the response is monotonic. A monotonic relationship supports the use of a parametric model such as the four-parameter logistic, which estimates the baseline response, the maximum effect, the slope, and the half-maximal effective dose (ED50). Non-monotonic relationships, such as a bell-shaped curve, require either a different model or a reconsideration of the underlying pharmacology, because they often indicate a secondary mechanism at higher doses or a receptor desensitization process. The choice between linear regression, nonlinear regression, and nonparametric methods depends on the endpoint type and the number of dose levels. Continuous endpoints with four or more dose levels can support nonlinear modeling. Binary endpoints, such as survival or the proportion of animals achieving a target response, are better analyzed with logistic regression, which yields a dose-response curve on the probability scale. The cecal ligation and puncture sepsis model reported by Barkhausen and colleagues used a 96-hour survival endpoint across four dose levels of two interventions, a design that allowed both the shape of the survival curve and the relative potency of the two compounds to be compared directly. That study also illustrates the value of including a vehicle control group and of randomising a sufficient number of animals per dose to detect differences in a high-mortality model. Model selection should be guided by the principle of parsimony. A simpler model that fits the data adequately is preferable to a more complex one, because complex models often produce unstable parameter estimates when the number of observations per dose is small. The Akaike information criterion provides a quantitative basis for comparing nested models. When the data do not support a parametric fit, nonparametric approaches such as the trimmed Spearman-Karber method can estimate the median effective dose without assuming a particular curve shape, at the cost of less precise estimates of the slope and maximum effect. ## Handling Variability and Missing Data Variability in veterinary dose-response studies arises from three principal sources: inter-animal pharmacokinetic differences, measurement error in the endpoint, and environmental or management factors. Inter-animal variability is often larger in outbred populations than in purpose-bred laboratory animals, and it is larger still in client-owned animals with concurrent disease. The statistical analysis should therefore include a random effect for animal if the design is repeated measures, or should at least test for heterogeneity of variance across dose groups. Heteroscedasticity is common in dose-response data because the variance often increases with the mean response, and it can be addressed by weighting the regression or by transforming the endpoint. Missing data are inevitable in studies involving client-owned animals. Owners may withdraw animals, animals may die from the disease or from treatment-related toxicity, and samples may be lost or mishandled. The analysis plan must specify how missing data will be handled before the study begins. Complete-case analysis is simple but biased if the missingness is related to the outcome. Multiple imputation or mixed-effects models that accommodate missing values under a missing-at-random assumption are preferable. For survival endpoints, the cecal ligation and puncture study demonstrates the standard approach: animals that die are censored at the time of death, and the analysis uses time-to-event methods instead of a simple binary outcome at a fixed time point. ## Documentation and Reporting Standards The reporting of dose-response studies should follow the ARRIVE guidelines, which specify the minimum information required for transparent and reproducible animal research. The guidelines require a clear description of the experimental unit, the randomisation procedure, the blinding method, the sample size calculation, and the statistical methods. They also require that all animals be accounted for, including those excluded from the analysis and the reasons for exclusion. The EQUATOR Network maintains a comprehensive library of reporting guidelines, and veterinary researchers should consult it for species-specific extensions such as REFLECT for livestock studies. The dose-response curve itself should be reported graphically with the fitted model and its confidence intervals, also as a table of point estimates. The parameters of the fitted model, including the ED50, the slope, and the maximum effect, should be reported with their standard errors. When the study is intended to support a regulatory submission, the analysis should be pre-specified in a statistical analysis plan that is finalised before the data are unblinded. Post-hoc analyzes may be reported as exploratory, but they should be clearly labelled as such and should not be used to support primary conclusions. ## Species-Specific Considerations The correct design choices differ across species in ways that go beyond simple scaling of doses. In food-producing animals, the primary endpoint may be a production parameter such as weight gain or feed conversion, and the study must account for group housing effects by using the pen or the flock as the experimental unit. In companion animals, the endpoint is more often a clinical score or an owner-reported outcome, and the study must account for the fact that the owner and the veterinarian may have different perceptions of treatment success. The MSD Veterinary Manual provides species-specific guidance on clinical pharmacology and on the interpretation of therapeutic responses, and it should be consulted when designing studies in less familiar species. The regulatory context also differs by species and by region. The World Organization for Animal Health sets international standards for the evaluation of veterinary products, and its terrestrial animal health code addresses the requirements for demonstrating efficacy and safety in target species. The American Veterinary Medical Association provides practice resources that address the conduct of clinical research in client-owned animals, including the ethical obligations to owners and to the animals themselves. Researchers should determine which standards apply to their jurisdiction and their species of interest before finalising the protocol. | Design choice | Preferred when | Avoid when | |---|---|---| | Parallel fixed-dose | Wide therapeutic index, chronic dosing, multiple dose levels feasible | Limited animal numbers, high inter-animal variability | | Dose titration | Narrow therapeutic index, chronic disease, individualised dosing expected | Acute disease, short treatment window, carryover effects | | Cross-over | Stable disease, short half-life, washout feasible | Food animals with long withdrawal periods, progressive disease | | Adaptive escalation | Rare conditions, limited subjects, toxicity is the main concern | Confirmatory efficacy is required, regulatory submission planned | | Two-stage pilot plus confirmatory | Uncertainty about dose range, novel drug class | Tight timeline, limited funding for two sequential phases | ## Recognized Complications and Failure Modes Dose-response studies in veterinary pharmacology fail in predictable patterns. The most consequential failure is dose-range misspecification, where all doses fall on the plateau of the response curve and the analysis returns a flat model. Detect this early by plotting interim response data against the log-dose axis. If the lowest dose produces near-maximal effect, the range is shifted too high. If no dose produces a measurable effect above control, the range is shifted too low or the endpoint is insensitive. Both findings justify protocol amendment before the study accumulates irreversible data loss. A second failure mode is carryover and period effects in crossover designs. When washout periods are too short, responses in later periods reflect residual drug from earlier periods. The discriminating check is a period-by-treatment interaction plot. Unequal carryover can also be detected by comparing responses in the first period only, which are uncontaminated. If period effects appear, restrict analysis to first-period data or apply a model that estimates carryover directly. A third failure mode is dose-confounded toxicity. When the highest dose produces systemic illness, the observed response conflates pharmacologic effect with toxicity. Weight loss, reduced food intake, and altered activity scores are early indicators. The check is to compare body weight trajectories and clinical examination findings across dose groups before analyzing the primary endpoint. If toxicity is present, the high-dose group may need exclusion from efficacy analysis, and the maximum tolerated dose becomes the study's upper boundary. A fourth failure mode is loss of blinding through visible drug effects. If the test article produces recognizable sedation, salivation, or injection-site reactions, personnel may infer treatment assignment. Use an active placebo that mimics the most visible adverse effect, or separate the team that assesses outcomes from the team that administers treatment. Formal blinding integrity checks, where raters guess treatment assignment, can quantify the problem. | Observation | Likely cause | Discriminating check | |---|---|---| | Flat response across all doses | Dose range misspecified | Plot interim responses, compare lowest dose to control | | Period-dependent responses | Carryover or period effects | Period-by-treatment interaction plot, first-period analysis | | High-dose group loses weight | Dose-confounded toxicity | Body weight trajectories, clinical examination findings | | Raters guess treatment correctly | Blinding failure from visible effects | Blinding integrity questionnaire, separate assessment team | ## Common Errors and Corrective Actions Less experienced investigators often select doses by rounding published values from other species without adjusting for allometric scaling, protein binding, or metabolic pathway differences. The corrective action is to perform a pilot study in the target species with three or four doses spanning at least one order of magnitude, then use those data to set the definitive range. Another frequent error is treating the dose-response curve as linear instead of log-linear. Analysis on the log-dose scale is standard because receptor occupancy and most pharmacodynamic models follow this relationship. Plotting raw dose against response will obscure the true curve shape. A third error is using too few animals per dose group while compensating with more dose levels. This spreads the sample too thin and reduces power at every point. The corrective action is to prioritize replication at the doses that define the curve's slope and plateau, typically the middle and upper doses, instead of distributing animals evenly. A fourth error is measuring the endpoint at a single time point when the drug's effect peaks at different times across doses. Higher doses may peak earlier or later than lower doses. Serial measurements with a predefined peak-detection rule, or an area-under-the-curve endpoint, avoid this bias. ## Limitations of Current Evidence The veterinary literature contains few formal dose-response studies compared with human medicine. Many published dose recommendations derive from anecdotal reports, small case series, or extrapolation from other species. The [MSD Veterinary Manual](https://www.msdvetmanual.com/) provides species-specific clinical guidance, but it does not substitute for rigorous dose-finding data in the target population. Regulatory standards for reporting animal research, such as the [ARRIVE guidelines](https://arriveguidelines.org/), have improved transparency, yet many older studies remain underpowered and poorly documented. Expert opinion still differs on several design questions. Whether to use fixed doses or body-weight-scaled doses remains contested, particularly for drugs with narrow therapeutic indices in growing animals. The choice between a no-observed-effect level and a minimum effective dose as the study's lower anchor also divides opinion. Some authorities favour including a toxicology-informed upper dose even when efficacy is the primary objective, while others argue this wastes animals and resources. These disagreements persist because comparative design studies in veterinary species are scarce. ## Referral, Consultation, and Reporting Dose-response studies that encounter unexpected toxicity, mortality, or lack of efficacy warrant specialist consultation. A veterinary clinical pharmacologist can review dose selection, sampling schedules, and analysis plans before the study begins. A veterinary pathologist should examine tissues from any animal that dies or is euthanised during the study, particularly at the highest dose. Laboratory involvement is required for assay validation, including proof of selectivity, recovery, and matrix effects for each species studied. Regulatory reporting obligations vary by jurisdiction and study purpose. Studies conducted for marketing authorisation must follow the standards of the relevant authority, and the [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) apply to studies involving animals destined for international trade. Adverse events that are unexpected, serious, or fatal should be reported to the appropriate pharmacovigilance system, and the [AVMA practice resources](https://www.avma.org/resources-tools) can direct investigators to applicable guidance. When in doubt about reporting obligations, consult the institutional animal care and use committee and the regulatory affairs office before the study starts, not after an adverse event occurs. ## Frequently Asked Questions ### How Many Dose Levels Can I Realistically Include With a Limited Budget? A minimum of three dose levels plus a control is required for any meaningful dose-response analysis. With tight resources, prioritize a well-spaced three-dose design over a poorly powered five-dose study. The ELLDOPA trial, a multicentre dosage-ranging study, used four treatment groups including placebo, demonstrating that a modest number of levels can yield definitive results when sample sizes per group are adequate [dose-ranging trial design in Parkinson's disease](https://pubmed.ncbi.nlm.nih.gov/16222436/). If only two active doses are feasible, consider a fixed-sequence design with a narrow dose range and accept that slope estimation will be imprecise. Pilot data from published literature or institutional archives can refine spacing without additional pilot animals. ### What Should I Do When the Recommended Equipment or Assay Platform Is Unavailable? Use the most sensitive validated assay you can access, even if it measures a surrogate endpoint. For behavioral or subjective outcomes, structured observer rating scales can substitute for automated systems, provided raters are blinded and trained to acceptable inter-rater reliability. The development of the Hallucinogen Rating Scale illustrates how carefully constructed instruments can capture graded drug effects when objective biomarkers are lacking [rating scale development for graded drug effects](https://pubmed.ncbi.nlm.nih.gov/8297217/). Document any assay substitution in the study protocol and report it under the ARRIVE guidelines, which require transparent description of all materials and methods [ARRIVE 2.0 reporting standards](https://arriveguidelines.org/). Validate the substitute against the reference method in a small pilot cohort before committing to the full study. ### How Do I Design a Dose-Response Study for a Food-Producing Species? Add a second dimension to the design: tissue residue depletion. Dose-response efficacy data must be paired with withdrawal period estimation at the highest proposed dose, because the approved withdrawal interval applies across the labelled dose range. Consult the [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) for international expectations on residue safety and trade implications. Include additional animals at the high dose for serial tissue sampling at multiple time points after the final dose. The efficacy endpoint and the residue endpoint may require different statistical models, so plan separate analyzes from the outset. Regulatory authorities in most jurisdictions require residue data generated under good laboratory practice, which increases cost and timeline compared with efficacy-only studies. ### What Records Must I Keep for a Dose-Response Study Beyond the Raw Data? Maintain a prospective audit trail covering dose preparation, administration times, batch numbers, and any deviation from the randomisation schedule. Record environmental conditions that could influence drug metabolism, such as temperature, photoperiod, and concurrent medication. The ARRIVE guidelines specify that publications must include details of housing, husbandry, and experimental procedures sufficient for replication [ARRIVE 2.0 reporting standards](https://arriveguidelines.org/). Keep version-controlled protocol amendments with dated signatures. Retain calibration certificates for all measurement instruments and logs of assay runs. These records support both regulatory inspection and later meta-analysis. Store electronic data in a format that remains readable after software upgrades, and archive paper records for at least the period required by your institution or funding body. ### How Should I Explain the Study Design to an Animal Owner or Production Manager? Frame the explanation around the purpose of dose selection: the study determines the lowest dose that achieves the target effect with the widest safety margin. Owners should understand that some animals receive a dose that may be subtherapeutic, and that this is deliberate and necessary for the scientific question. Explain the randomisation and blinding in plain terms, emphasizing that neither you nor they will know the allocation during the study. Provide a written summary of the expected time commitment, observation schedule, and any anticipated adverse effects. The [AVMA practice resources](https://www.avma.org/resources-tools) include guidance on client communication and informed consent for research participation. Reassure owners that withdrawal from the study is possible at any time without penalty to their animal's care. ### When Should I Use a Pilot Study Instead of a Full Dose-Response Design? Use a pilot study when the dose range is uncertain, the endpoint variability is unknown, or the formulation has not been tested in the target species. A pilot with six to ten animals per group can establish feasibility, refine the dose range, and generate variance estimates for sample size calculation. The sepsis model study cited earlier used a two-stage approach: a 96-hour survival study followed by a 24-hour mechanistic study, allowing dose selection to inform the second phase [dose-response design in a murine sepsis model](https://pubmed.ncbi.nlm.nih.gov/21336117/). Pilot data also reveal unexpected toxicity at high doses that would otherwise compromise the main study. Report pilot results separately and do not pool them with the confirmatory data, as the pilot's purpose is design refinement, not hypothesis testing. ## Related Clinical & Scientific Guides * [Conducting Systematic Reviews of Veterinary Diagnostic Test Accuracy](/knowledge/veterinary-medicine/veterinary-research-methods/conducting-systematic-reviews-veterinary-diagnostic-test-accuracy) * [Bias in Veterinary Research: Types, Sources, and Mitigation](/knowledge/veterinary-medicine/veterinary-research-methods/bias-veterinary-research-types-sources-mitigation) * [Cluster Randomized Trials in Veterinary Research: Design and Analysis](/knowledge/veterinary-medicine/veterinary-research-methods/cluster-randomized-trials-veterinary-research-design-analysis) ## References and Further Reading - [Growth factors and treatment of intervertebral disc degeneration.](https://pubmed.ncbi.nlm.nih.gov/15564925/). 2004. - [Dose-response study of N,N-dimethyltryptamine in humans. II. Subjective effects and preliminary results of a new rating scale.](https://pubmed.ncbi.nlm.nih.gov/8297217/). 1994. - [Selective blockade of interleukin-6 trans-signaling improves survival in a murine polymicrobial sepsis model.](https://pubmed.ncbi.nlm.nih.gov/21336117/). 2011. - [Odd Chain Fatty Acids, New Insights of the Relationship Between the Gut Microbiota, Dietary Intake, Biosynthesis and Glucose Intolerance.](https://pubmed.ncbi.nlm.nih.gov/28332596/). 2017. - [Does levodopa slow or hasten the rate of progression of Parkinson's disease?](https://pubmed.ncbi.nlm.nih.gov/16222436/). 2005. - [ARRIVE Guidelines 2.0 for Reporting Animal Research](https://arriveguidelines.org/). PLOS Biology, 2020. - [EQUATOR Network Reporting Guidelines](https://www.equator-network.org/). EQUATOR Network. - [MSD Veterinary Manual, Professional Edition](https://www.msdvetmanual.com/). MSD Veterinary Manual. - [American Veterinary Medical Association Practice Resources](https://www.avma.org/resources-tools). American Veterinary Medical Association. ## Related Articles - [Designing Questionnaire Studies for Veterinary Research](/knowledge/veterinary-medicine/veterinary-research-methods/designing-questionnaire-studies-veterinary-research) - [Cohort Studies in Veterinary Research: Design and Interpretation](/knowledge/veterinary-medicine/veterinary-research-methods/cohort-studies-veterinary-research-design-interpretation) - [Evaluating Prognostic Factor Studies in Veterinary Medicine](/knowledge/veterinary-medicine/veterinary-research-methods/evaluating-prognostic-factor-studies-veterinary-medicine) - [Observational Studies in Veterinary Medicine: Strengths and Weaknesses](/knowledge/veterinary-medicine/veterinary-research-methods/observational-studies-veterinary-medicine-strengths-weaknesses) - [Conducting Pharmacovigilance Studies in Veterinary Medicine](/knowledge/veterinary-medicine/veterinary-research-methods/conducting-pharmacovigilance-studies-veterinary-medicine) > 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.