# Implementing Adaptive Trial Designs in Veterinary Clinical Research


## Key Takeaways

- Adaptive trial designs allow for pre-specified modifications to study elements (e.g., sample size, treatment allocation) based on accumulating data, enhancing efficiency and ethical considerations by potentially reducing the number of animals exposed to ineffective treatments, particularly valuable in veterinary medicine where enrollment can be slow and costly.
- Implementing adaptive designs necessitates robust statistical infrastructure, including pre-defined statistical analysis plans detailing interim analysis timing, stopping rules (for efficacy or futility), and methods for Type I error control (e.g., alpha-spending functions), alongside a dedicated data monitoring committee with expertise in adaptive methods and species-specific knowledge.
- Sample size re-estimation, a key adaptive feature, allows adjustments based on interim data to correct for uncertain nuisance parameters like variance or control event rates, with blinded re-estimation generally preferred in veterinary trials to preserve Type I error control, though unblinded re-estimation may be considered for more complex adaptations.
- Multi-arm adaptive designs enable the dropping of poorly performing treatment arms during the trial, concentrating resources on more promising regimens, and require careful statistical adjustments (e.g., Dunnett-type adjustments) to maintain overall Type I error rate, with protocol specification of allocation ratios post-dropping being critical.
- Response-adaptive randomization, where treatment assignment probabilities shift towards better-performing arms, is ethically attractive for reducing exposure to inferior treatments but requires rapid outcome ascertainment and robust central randomization capabilities, making it less suitable for veterinary trials with slow outcome measurements.
- Rigorous documentation and reporting are paramount, adhering to guidelines like ARRIVE 2.0 and EQUATOR Network, with every adaptation justified by pre-specified rules, not ad hoc decisions, to ensure transparency, reproducibility, and regulatory compliance.

---

Adaptive trial designs modify prespecified elements of a study while data accumulate, using interim analyzes to inform changes without compromising validity. This article provides veterinary researchers with the procedural framework needed to implement these designs in clinical studies across species. It covers the scientific rationale, operational planning, interim analysis conduct, sample size re-estimation, and reporting requirements, with emphasis on the practical decisions that distinguish a well-run adaptive trial from a protocol that has drifted from its statistical foundations.

The intended reader is a veterinary researcher with working knowledge of clinical trial methodology who is considering an adaptive design for a companion animal, livestock, or wildlife study. The article answers a specific procedural question: what must be planned, documented, and executed before the first interim analysis, and how do those requirements differ from a fixed-design trial? Bayesian-specific methods are excluded, the focus is on frequentist group-sequential and sample size re-estimation approaches that can be implemented with standard statistical software and trial infrastructure.

## At a Glance

| Parameter | Consideration |
|---|---|
| Design selection | Choose the adaptive element before enrollment, specify in protocol with statistical analysis plan |
| Interim analyzes | Predefine timing, stopping rules, and the data monitoring committee charter |
| Sample size re-estimation | Use blinded or unblinded approaches, specify the decision rule and timing |
| Type I error control | Required for confirmatory trials, use alpha-spending functions or other prespecified methods |
| Statistical infrastructure | Adaptive trials require additional statistician time and data management resources |
| Regulatory and reporting | Follow ARRIVE 2.0 and EQUATOR Network guidelines for transparency |
| Species considerations | Account for spontaneous disease, owner compliance, and production setting constraints |
| Documentation | Every adaptation must be justified by prespecified rules, not by observed results alone |

## The Logic of Adaptive Design in Veterinary Research

Adaptive designs address a structural inefficiency in fixed-design trials: the sample size, endpoints, and treatment allocation are locked before any data exist, even though the trial itself will generate information that could refine those choices. In veterinary medicine, where patient enrollment is often slower and more expensive than in human trials, the ability to stop early for futility or efficacy, or to adjust the sample size when the observed effect differs from the assumed effect, can conserve resources and reduce the number of animals exposed to inferior treatments.

The statistical foundation rests on the principle that adaptations must be prespecified and must not introduce bias. A trial that changes its sample size because the interim results look promising, without a rule that was written before those results were seen, loses its error control properties. The trial design literature distinguishes between adaptations that are planned in the protocol and those that are ad hoc responses to data. Only the former preserve the frequentist guarantees that allow confirmatory conclusions.

Veterinary trials face particular constraints that make adaptive designs attractive. Spontaneous disease models in client-owned animals have variable baseline severity, owner adherence affects treatment response, and production animals are subject to management decisions that can confound outcomes. The adaptive framework accommodates this variability by allowing the trial to respond to observed variance and effect sizes instead of relying on estimates from prior studies that may not transfer across breeds, housing systems, or geographic regions.

## Trial Infrastructure and Resource Planning

Adaptive designs require more than a modified protocol. The operational infrastructure must support rapid data cleaning, timely interim analyzes, and secure communication of results to a monitoring committee. A costing exercise conducted across seven clinical trials units in the United Kingdom found that estimated resources for adaptive trials were consistently higher than for fixed designs, with the median increase attributable to additional statistician time, data management, and trial management effort [Costs and staffing resource requirements for adaptive clinical trials](https://pubmed.ncbi.nlm.nih.gov/34696781/). Veterinary research groups should budget for these costs explicitly instead of assuming that an adaptive protocol will reduce overall workload.

The data monitoring committee, sometimes called a data safety monitoring board, is a structural requirement for unblinded interim analyzes. This committee must include a statistician with adaptive design expertise, a clinician with species-specific knowledge, and ideally a member with ethics or welfare expertise. The committee charter must specify who sees unblinded data, how decisions are communicated to the sponsor, and what happens if the committee cannot reach consensus. In veterinary trials, the charter should also address animal welfare stopping rules that may operate independently of statistical stopping rules.

### Interim Analysis Timing

The timing of interim analyzes must balance statistical efficiency against operational feasibility. Early interim analyzes, conducted after a small fraction of the target sample size has been enrolled, can stop a trial quickly if the treatment effect is large or absent, but they provide imprecise estimates of variance and effect size. Later interim analyzes offer more information but commit more animals to the trial before any adaptation occurs. The protocol must specify the information fraction at each analysis, expressed as the proportion of the final target sample size or the proportion of the total planned information, not as a calendar date.

## Group-Sequential Designs in Veterinary Trials

Group-sequential designs divide enrollment into stages, with an interim analysis after each stage and prespecified stopping boundaries for efficacy, futility, or both. The alpha-spending approach, which allocates the overall type I error across the interim analyzes, allows the timing of analyzes to deviate from the original plan without inflating the error rate. This flexibility matters in veterinary trials because enrollment often proceeds more slowly or more quickly than projected.

A veterinary example illustrates the practical value. A randomized, blinded, placebo-controlled trial of medical-grade honey for canine nasal intertrigo used an adaptive group-sequential design and rejected the null hypothesis at the first interim analysis, with the placebo outperforming the honey on cytological and clinical composite scores [Medical honey for canine nasal intertrigo](https://pubmed.ncbi.nlm.nih.gov/32760092/). The design allowed the investigators to stop early and avoid enrolling additional dogs onto a treatment that was performing worse than control. This case demonstrates that adaptive designs are also for showing superiority, they can protect animals from ineffective or harmful interventions.

### Futility Stopping

Futility stopping rules are particularly valuable in veterinary research because they reduce the number of animals exposed to a treatment that cannot plausibly demonstrate benefit. A conditional power calculation, performed at the interim analysis, estimates the probability that the final analysis will reach significance given the observed data and the assumed effect size for the remainder of the trial. If this probability falls below a prespecified threshold, the trial stops for futility. The threshold must be chosen with care: a threshold that is too high stops promising trials prematurely, while a threshold that is too low commits animals to a futile study.

## Sample Size Re-Estimation in Veterinary Trials

Sample size re-estimation allows the trial to adjust its projected enrollment target based on data accumulated at an interim analysis. The most common trigger for re-estimation is a nuisance parameter, such as the variance of the primary outcome or the event rate in the control arm, which was uncertain at the design stage. Veterinary trials frequently face this uncertainty because pilot data are scarce, and published estimates often come from different breeds, ages, or management systems.

Two broad approaches exist. Blinded re-estimation uses only the pooled data across treatment arms, so the treatment effect estimate is not revealed and the type I error rate is largely protected. Unblinded re-estimation uses the observed between-arm difference, which requires an adjustment to the final analysis to preserve the nominal significance level. For veterinary researchers, blinded re-estimation is often the more defensible choice when the goal is simply to correct an underpowered design. Unblinded re-estimation becomes attractive when the trial must also decide whether to drop an arm or change the primary endpoint, but it demands greater statistical oversight.

The decision to re-estimate should be specified in the protocol before enrollment begins. The protocol must state the timing of the re-estimation, the parameter to be estimated, the range of sample sizes that will be considered, and the rule that maps the observed estimate to a revised sample size. A common rule is to increase the sample size only if the observed effect size falls within a pre-specified promising zone, and to stop for futility if it falls below that zone. This prevents the trial from continuing when the treatment effect is too small to be clinically meaningful, and it prevents an unbounded increase in cost when the effect is merely modest.

The [Costing Adaptive Trials project](https://pubmed.ncbi.nlm.nih.gov/34696781/) found that adaptive trials consistently required more statistical and data management time than fixed designs, with the median increase estimated across multiple trial units. Sample size re-estimation was among the design features that drove those additional costs. Veterinary researchers should therefore budget for additional statistician time and for the possibility of a longer enrollment window, even when the final sample size is smaller than the fixed-design alternative.

## Treatment Arm Selection and Dropping

Multi-arm adaptive designs allow a trial to begin with several candidate treatments and drop those that perform poorly at interim analyzes. This structure suits veterinary medicine well, where multiple formulations, dose levels, or adjunctive therapies may compete for the same patient population. The platform can continue with the remaining arms, and the final analysis includes only the treatments that survived the selection process.

The statistical penalty for arm dropping depends on the number of arms, the timing of the interim analyzes, and the correlation between the test statistics. Designs that use a Dunnett-type adjustment, where each active arm is compared against a shared control, preserve the family-wise error rate more efficiently than designs that treat each comparison as independent. The protocol must specify the allocation ratio after an arm is dropped, because the control arm may need additional enrollment to maintain power for the remaining comparisons.

Veterinary researchers should consider whether the dropped arm's data will still be reported. Even when an arm is stopped for futility, the data collected from that arm contribute to the safety database and to secondary analyzes. The [adaptive trial designs described for tuberculosis drug development](https://pubmed.ncbi.nlm.nih.gov/25946350/) illustrate this principle, where early-phase selection of drug combinations allows limited patient resources to be concentrated on the most promising regimens. The same logic applies in veterinary medicine, particularly in food animals where per-animal trial costs are high and enrollment windows are tied to production cycles.

## Response-Adaptive Randomisation

Response-adaptive randomisation changes the probability of assignment to each treatment arm based on accumulating outcome data, so that more animals receive the treatment that is performing better. This design is ethically attractive in veterinary trials because it reduces the number of animals exposed to an inferior treatment. It is also statistically efficient when the outcome is measured quickly relative to the enrollment rate, because the adaptation can respond to a substantial fraction of the data.

The practical constraints are significant. Response-adaptive randomisation requires a central randomisation service that can update allocation probabilities in real time, which may not be available in multi-site veterinary trials. It also creates a risk of operational bias if site staff can infer the treatment allocation from the changing randomisation ratio. The [HIV vaccine adaptive trial literature](https://pubmed.ncbi.nlm.nih.gov/21508308/) notes that adaptive designs can accelerate the screening of poor candidates, but the same source emphasizes that the choice of design must match the speed of the outcome measurement and the capacity for central coordination.

For veterinary trials with slow outcome ascertainment, such as those measuring survival to discharge or cure at 28 days, response-adaptive randomisation is rarely the right choice. The adaptation would occur too late to influence most allocations. Group-sequential designs with futility stopping are usually more practical, because they require only periodic analyzes instead of continuous updating.

## Documentation and Reporting Standards

Adaptive trials place a heavy burden on documentation, because every design decision must be traceable to a pre-specified rule. The protocol must include a statistical analysis plan that describes the adaptive elements, the timing of each interim analysis, the decision rules, and the method for adjusting the final analysis. Any deviation from the pre-specified rules must be reported and justified, because unplanned adaptations can inflate the type I error rate and undermine the credibility of the results.

Reporting should follow the [ARRIVE guidelines for animal research](https://arriveguidelines.org/), which specify the minimum information required for transparent and reproducible publications. The [EQUATOR Network reporting guideline library](https://www.equator-network.org/) provides additional resources, including the CONSORT extension for adaptive designs, which is the relevant standard for randomised trials. Veterinary journals increasingly expect compliance with these standards, and funding bodies may require evidence of reporting quality at the grant stage.

The final report should state the number of interim analyzes actually performed, the results of each, and the decisions taken. It should also report the final sample size against the planned range, and explain any discrepancy. This transparency allows readers to assess whether the adaptive elements were conducted as intended and whether the conclusions are robust to the design choices.

## Regulatory and Ethical Review Considerations

Regulatory bodies and animal ethics committees evaluate adaptive designs differently from fixed designs. The key question is whether the adaptation introduces a risk to animal welfare or to the scientific validity of the trial. A well-specified adaptive design that reduces the number of animals exposed to an ineffective treatment is generally viewed favourably, because it aligns with the ethical principle of minimizing animal use.

The [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) provide the international framework for animal health research and disease control, and they should be consulted when the trial involves food animals or notifiable diseases. Regional requirements vary, and the [AVMA professional practice resources](https://www.avma.org/resources-tools) can help identify relevant guidance for companion animal trials in North America. The [MSD Veterinary Manual](https://www.msdvetmanual.com/) offers species-specific clinical context that may inform the choice of endpoints and the interpretation of interim results.

The ethics application must describe the adaptive elements in plain language, including the circumstances under which the trial may stop early or change the allocation ratio. The committee will want assurance that the adaptation cannot compromise animal welfare, for example by extending the trial when a treatment is clearly harmful. The protocol should therefore include a safety monitoring plan that operates independently of the adaptive decision rules, with explicit criteria for stopping the trial for harm.

A practical checklist for regulatory submission includes the following elements: the pre-specified decision rules, the statistical methods for preserving the type I error rate, the safety monitoring plan, the plan for reporting deviations, and the justification for the chosen adaptive design over a fixed design. The [Costing Adaptive Trials project](https://pubmed.ncbi.nlm.nih.gov/34696781/) found that the additional regulatory and documentation burden was a major driver of the increased cost of adaptive trials, so this checklist should be prepared early and reviewed by a statistician before submission.

| Design Element | When to Use | Key Requirement | Common Failure Mode |
|---|---|---|---|
| Group-sequential with futility stopping | Outcome measured within weeks, enrollment steady | Pre-specified stopping boundaries | Boundaries set too wide, so trial continues despite negligible effect |
| Sample size re-estimation | Variance or control event rate uncertain | Blinded estimation preferred | Unblinded estimation without type I error adjustment |
| Arm dropping | Multiple candidate treatments, limited patient pool | Dunnett-type adjustment for multiplicity | Dropping an arm based on a non-significant but clinically relevant difference |
| Response-adaptive randomisation | Rapid outcome, central randomisation capacity | Real-time allocation updates | Slow outcome measurement makes adaptation ineffective |

## Recognized Complications and Failure Modes

Adaptive designs fail in characteriztic ways, and most failures are detectable before they compromise the trial. The most consequential failure mode is inflation of the type I error rate through unplanned interim peeking. A group-sequential design that specifies three interim analyzes loses its error control if an additional, unplanned analysis is performed and the stopping boundaries are not adjusted. Detection requires an audit trail that records every look at the data, including unblinded summaries, and a statistical analysis plan that pre-specifies the number and timing of all analyzes.

Operational bias arises when interim results leak to investigators, enrolling clinicians, or animal caretakers. In veterinary trials, the risk is heightened because the same individual may enrol animals, administer treatment, and assess outcomes. Early detection depends on blinding checks, separation of the data monitoring committee from the study team, and documentation of who accessed unblinded reports. A second failure mode is drift in the study population over time. If response-adaptive randomisation allocates more animals to a superior arm, the case mix may shift as the trial progresses, and the final estimate can reflect a population that differs from the target population. Monitoring baseline characteriztics by randomisation wave and comparing early versus late enrollments will expose this drift.

Sample size re-estimation can fail when the blinded estimate of the nuisance parameter is biased. For example, if the overall event rate is used to re-estimate sample size in a superiority trial, treatment effect heterogeneity can inflate the estimate. The discriminating check is to compare the blinded estimate against the unblinded interim estimate prepared by the data monitoring committee. Divergence beyond the pre-specified tolerance indicates that the re-estimation procedure requires correction or that the underlying assumptions about the control arm are wrong.

## Common Errors and Corrective Actions

Less experienced trialists frequently confuse the roles of the interim analysis and the final analysis. The interim analysis is not a preliminary test of efficacy, it is a decision point governed by pre-specified boundaries. A common error is to stop for efficacy at an interim analysis using a nominal p-value that does not account for the repeated looks. The corrective action is to use the spending function or alpha-spending approach specified in the analysis plan, which preserves the overall error rate across all analyzes.

Another frequent error is the failure to pre-specify the decision rules for dropping an arm or stopping for futility. When the rules are written after the interim data are seen, the design is no longer adaptive, it is ad hoc. The corrective action is to draft the complete decision matrix, including the thresholds for each action, before enrollment begins and to lodge it with the data monitoring committee charter. A related error is the use of a futility stopping rule that is too lenient, which stops a promising arm early, or too strict, which wastes resources on an ineffective arm. The boundaries should be calibrated to the clinically meaningful effect size, not to statistical significance alone.

A third error is the failure to plan for the logistical consequences of an interim decision. If an arm is dropped, the randomisation schedule must be updated, the case report forms revised, and the animal care staff informed without revealing the reason. Trials that lack a pre-specified operational plan for these transitions experience delays and protocol deviations. The corrective action is to rehearse the interim decision process with the study team before the trial starts, including the communication scripts and the timeline for implementing the decision.

## Limitations of the Current Evidence

The evidence base for adaptive designs in veterinary medicine is thin. Most methodological guidance derives from human clinical trials, where the regulatory environment, the outcome measures, and the ethical framework differ from veterinary practice. The tuberculosis and HIV vaccine literature demonstrates the value of adaptive designs in human infectious disease trials, but the extrapolation to veterinary species is limited by differences in disease models, endpoint validation, and the absence of a veterinary regulatory pathway equivalent to human drug approval processes. The canine nasal intertrigo trial provides a rare veterinary example of a group-sequential design, but it is a single study in a single condition, and its findings cannot be generalized to other species or therapeutic areas.

Expert opinion still differs on several points. The optimal timing of interim analyzes in veterinary trials remains contested. Some methodologists argue for early interim analyzes to minimize the number of animals exposed to inferior treatments, while others caution that early analyzes are unstable because the outcome data are incomplete and the variance estimates are imprecise. There is also disagreement about the use of response-adaptive randomisation in small trials. Proponents argue that it reduces the number of animals on inferior arms, while critics note that the operating characteriztics are poorly understood when the sample size is below 50 per arm and the outcome is measured with error.

The resource implications of adaptive designs are better documented. The Costing Adaptive Trials project found that the estimated resources for an adaptive trial were consistently higher than for a fixed design, with the median increase driven by additional statistical support, data management, and trial management time. Veterinary researchers should budget for these additional costs, particularly when the trial is conducted within a university or practice-based setting where statistical support may be limited.

## Referral, Consultation, and Reporting

Adaptive designs require specialised statistical expertise. A veterinary researcher who does not have access to a statistician experienced in group-sequential methods should consult one before the protocol is finalised. The consultation should cover the choice of design, the specification of the analysis plan, and the calibration of the stopping boundaries. A statistician should also be involved in the interim analyzes, because the decision rules are only valid if the computations are performed correctly.

Laboratory involvement is required when the interim analysis depends on biomarkers or surrogate endpoints that are measured in a central laboratory. The laboratory must be able to return results within the time window specified in the protocol, and the assay performance must be stable across the trial. If the assay has high batch-to-batch variability, the interim analysis may be driven by laboratory noise instead of treatment effect. The corrective action is to include assay quality control samples in each batch and to monitor the coefficient of variation across batches.

Regulatory reporting is required when an interim decision changes the course of the trial in a way that affects animal welfare or the validity of the data. In jurisdictions where veterinary clinical trials are overseen by an ethics committee or an animal welfare body, the committee must be informed of any protocol amendment that results from an interim analysis. The reporting should include the rationale for the decision, the data that supported it, and the planned amendments. The ARRIVE guidelines and the EQUATOR Network resources provide frameworks for reporting the design and the interim decisions transparently, and they should be consulted when preparing the final manuscript.

| Observation | Likely cause | Discriminating check |
|---|---|---|
| Type I error exceeds nominal level | Unplanned interim analyzes | Audit trail of all data looks, compare planned versus actual analysis count |
| Interim results predict final results too closely | Unblinding or operational bias | Blinding checks, review who accessed unblinded reports |
| Baseline characteriztics shift across enrollment waves | Response-adaptive randomisation changing case mix | Compare baseline variables by randomisation wave |
| Blinded sample size re-estimation diverges from unblinded estimate | Nuisance parameter bias | Compare blinded versus unblinded estimates at interim |
| Arm dropped but enrollment continues to that arm | Logistical failure in implementing interim decision | Verify randomisation schedule update and staff communication records |
| Interim analysis delayed by laboratory turnaround | Assay or logistics bottleneck | Monitor laboratory turnaround time against protocol window |

## Frequently Asked Questions

### How Much More Expensive Is an Adaptive Trial Compared to a Fixed Design?

Adaptive trials require more statistical, data management, and trial coordination time than fixed designs. A costing exercise across seven clinical trials units found that estimated resources were always moderately higher for the adaptive version, with a median increase between 10% and 20% depending on design complexity. The additional cost is concentrated in the planning phase, where simulation work and pre-specified decision rules demand more statistician time, and in the conduct phase, where interim analyzes require scheduled data cleaning and unblinded review. Group-sequential designs with one interim analysis are the least costly option. Response-adaptive randomisation and multi-arm designs with dropping arms carry the highest premium. Budget for these increases explicitly in grant applications instead of absorbing them from other trial activities.

### What Should We Do When We Cannot Afford Dedicated Adaptive Trial Software?

Standard statistical software can support many adaptive designs if the team is disciplined about pre-specification. Group-sequential boundaries and sample size re-estimation formulas are implemented in common packages, and simulation for operating characteriztics can be run with general-purpose programming. The critical requirement is not specialised software but a pre-specified analysis plan that fixes the decision rules, the timing of interim analyzes, and the exact test statistics to be used. A statistician with experience in sequential methods is more valuable than any software purchase. If such a statistician is unavailable, consider simplifying the design to a single interim analysis for futility only, which requires less simulation infrastructure than designs with sample size re-estimation or arm dropping. Document all simulation code and decision rules in the trial master file.

### How Do Adaptive Designs Work in Small Animal Practice Compared to Livestock Trials?

The practical constraints differ substantially. In companion animal practice, patient recruitment is often slower and the case mix more heterogeneous, so interim analyzes may need to be triggered by calendar time instead of by a fixed number of events. The outcome measures are frequently owner-reported scales, which have higher variability and require larger sample sizes to detect the same effect size. In livestock trials, group housing and herd-level randomisation create clustering effects that must be incorporated into the design and the interim analysis plan. Production animals also allow repeated sampling and objective outcomes such as weight gain or lesion scores, which can reduce measurement error. The adaptive machinery itself is transferable across species, but the operating characteriztics must be simulated under the specific recruitment rate, outcome distribution, and clustering structure of the target population.

### What Records Must We Keep for an Adaptive Trial to Pass Regulatory or Ethical Scrutiny?

The trial master file must contain the full pre-specified protocol, including the adaptive elements and the statistical analysis plan with all decision rules. Every interim analysis requires a dated report documenting the unblinded results, the pre-specified decision rule applied, and the resulting action. The independent data monitoring committee must have terms of reference that define its membership, voting procedures, and communication lines to the sponsor. All simulation work used to establish operating characteriztics should be archived with version control. Any deviation from the pre-specified plan, however minor, must be documented with justification. Reporting standards such as the ARRIVE guidelines for animal research and the EQUATOR network's reporting resources provide checklists that help ensure the adaptive elements are described transparently in the final publication.

### How Do We Explain an Adaptive Design to an Institutional Animal Care and Use Committee?

Frame the design around animal welfare and scientific validity. Explain that adaptive elements such as futility stopping reduce the number of animals exposed to an ineffective treatment, while sample size re-estimation protects the trial from being underpowered, which would waste all animals used. Emphasize that the decision rules are fixed in advance and that an independent committee, not the investigators, reviews interim data. Provide the committee with the simulation results showing the expected distribution of final sample sizes and the probability of early stopping. The committee will want reassurance that the design does not allow unplanned peeking at accumulating data, which would inflate the type I error rate. A clear diagram of the decision tree and a plain-language summary of the stopping rules are usually sufficient.

### What Is the Minimum Sample Size Needed to Justify an Adaptive Design?

There is no fixed threshold, but adaptive designs become more attractive as the cost per animal and the uncertainty about the effect size increase. A group-sequential design with a single futility interim analysis can be useful even with a planned total sample size of 40 to 60 animals, particularly in rare diseases or when pilot data are weak. Designs with sample size re-estimation require a larger planned maximum because the re-estimation procedure needs enough interim data to estimate the variance reliably. The decision should be based on simulation: if the adaptive design reduces the expected sample size under the null or a small effect while preserving power under the target effect, it is justified. If the trial is large and the effect size is well established from prior work, a fixed design is simpler and cheaper.

## 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

- [Costs and staffing resource requirements for adaptive clinical trials: quantitative and qualitative results from the Costing Adaptive Trials project.](https://pubmed.ncbi.nlm.nih.gov/34696781/). 2021.
- [HIV-1 vaccines and adaptive trial designs.](https://pubmed.ncbi.nlm.nih.gov/21508308/). 2011.
- [Adaptive clinical trials in tuberculosis: applications, challenges and solutions.](https://pubmed.ncbi.nlm.nih.gov/25946350/). 2015.
- [Overcoming dengue vaccine challenges through next-generation virus-like particle immunization strategies.](https://pubmed.ncbi.nlm.nih.gov/40589870/). 2025.
- [The Yin and the Yang of Transformative Research During the COVID-19 Pandemic-A Perspective.](https://pubmed.ncbi.nlm.nih.gov/34249804/). 2021.
- [Medical honey for canine nasal intertrigo: A randomized, blinded, placebo-controlled, adaptive clinical trial to support antimicrobial stewardship in veterinary dermatology.](https://pubmed.ncbi.nlm.nih.gov/32760092/). 2020.
- [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.

## Related Articles

- [Pragmatic vs Explanatory Trials in Veterinary Clinical Research](/knowledge/veterinary-medicine/veterinary-research-methods/pragmatic-vs-explanatory-trials-veterinary-clinical-research)
- [Designing Adaptive Clinical Trials for Veterinary Medicine](/knowledge/veterinary-medicine/veterinary-research-methods/designing-adaptive-clinical-trials-veterinary-medicine)
- [Cross-Over Designs in Veterinary Clinical Trials: Advantages and Limitations](/knowledge/veterinary-medicine/veterinary-research-methods/cross-over-designs-veterinary-clinical-trials-advantages-limitations)
- [Blinding in Veterinary Clinical Research: Methods and Challenges](/knowledge/veterinary-medicine/veterinary-research-methods/blinding-veterinary-clinical-research-methods-challenges)
- [Handling Missing Data in Veterinary Clinical Research](/knowledge/veterinary-medicine/veterinary-research-methods/handling-missing-data-veterinary-clinical-research)

> 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.