# Animal Model Selection for Neurological Research ## Key Takeaways - Model selection for neurological research necessitates a precise alignment between the research question and the chosen animal model's capacity to reproduce specific disease features, such as aetiology, pathology, or symptomatology, rather than defaulting to conventional preparations. - Three critical validity metrics: face, construct, and predictive validity, must be rigorously assessed; a model may exhibit strong resemblance to human symptoms (face validity) but lack congruence with the underlying pathogenic mechanism (construct validity) or fail to predict therapeutic efficacy (predictive validity). - Pharmacological and toxin models are useful for investigating acute molecular insults or screening interventions targeting specific pathways but often fail to recapitulate the progressive, multifactorial nature of chronic neurological disorders. - Genetic models offer high construct validity for monogenic disorders by replicating known disease-associated genotypes, but their phenotypes can be subtle, variable, or require extended observation periods to emerge. - Species choice, ranging from rodents to non-human primates, must be driven by the biological question and translational goals, considering factors like brain complexity, physiological resemblance, and available genetic tools, while always prioritizing animal welfare and regulatory compliance. - Comprehensive baseline characterization, including sex as a biological variable, and a pre-defined monitoring plan with clear humane endpoints are essential for ensuring data integrity, interpreting results accurately, and upholding ethical standards throughout the research process. --- Neurological research spans a broad spectrum of questions, from molecular mechanisms of neurodegeneration to the behavioral consequences of network dysfunction. The choice of animal model determines which aspects of a human condition are reproduced, which experimental manipulations are feasible, and how confidently findings can be translated back to clinical populations. This article provides a decision framework for veterinary researchers selecting models for neurological studies, with emphasis on matching model features to the specific research question instead of defaulting to familiar or historically common preparations. The reader is assumed to be a veterinary scientist or graduate student designing a study involving laboratory animals. The article addresses the conceptual basis of model selection, the major categories of neurological disease models, species-specific considerations, and the practical constraints imposed by animal welfare oversight and translational goals. It does not cover models for pain, infectious disease, or other non-neurological research areas. ## At a Glance | Parameter | Decision Point | Reference | |---|---|---| | Research question type | Mechanism, therapy screening, biomarker discovery, or behavioral phenotyping | Study hypothesis | | Disease feature priority | Aetiology, pathology, motor signs, non-motor signs, or all combined | [Animal models of Parkinson's disease guide](https://pubmed.ncbi.nlm.nih.gov/34956652/) | | Model category | Pharmacological, toxin, genetic, surgical, or spontaneous | Disease-specific literature | | Species choice | Rodent, large animal, or non-human primate | Facility capacity and translational target | | Outcome measure validity | Face, construct, and predictive validity of the chosen readout | Published validation studies | | Welfare burden | Severity classification and refinement opportunities | [NC3Rs resources](https://www.nc3rs.org.uk/) | | Regulatory compliance | Institutional animal care and use approval, national standards | [Guide for the Care and Use of Laboratory Animals](https://grants.nih.gov/grants/olaw/guide-for-the-care-and-use-of-laboratory-animals.pdf) | ## The Logic of Model Selection A neurological disease model is an experimental system that reproduces selected features of a human or veterinary condition. No single model reproduces every aspect of a disease, and the selection process is therefore an exercise in prioritization. The researcher must first define which disease features are essential for the study. For Parkinson's disease, for example, the features include α-synuclein pathology, nigrostriatal dopaminergic degeneration, motor dysfunction, and a range of non-motor signs. A toxin model such as 6-hydroxydopamine reproduces the dopaminergic lesion and motor deficits but does not generate α-synuclein aggregates, whereas certain genetic models produce proteinopathy without robust cell loss. The review by Lama and colleagues organizes available Parkinson's models according to which disease features each one replicates, providing a practical template for this kind of feature-based selection. Three forms of validity guide the evaluation of a model. Face validity refers to the resemblance between the model phenotype and the human condition. Construct validity concerns whether the model is based on the same pathogenic mechanism as the disease. Predictive validity describes whether the model responds to therapeutic interventions in a way that forecasts clinical efficacy. A model may score highly on one form of validity and poorly on another. A kindling model of epilepsy, for instance, has strong face validity for focal seizures and excellent predictive validity for many antiseizure drugs, but its construct validity depends on whether epileptogenesis in the kindled brain truly mirrors the human disease process. Wang and colleagues note that the kindling model was historically the most widely used epilepsy preparation, yet newer models have been developed to mimic different seizure types, each with its own profile of strengths and limitations. ## Disease Mechanisms and Model Categories ### Pharmacological and Toxin Models Pharmacological models induce acute or subacute neurological dysfunction through administration of a drug or toxin. These preparations are valuable for studying the consequences of a defined molecular insult and for screening interventions that target the affected pathway. Their principal limitation is that they reproduce a single pathogenic event instead of the full disease process. Toxin models of Parkinson's disease, for example, create a dopaminergic lesion that mimics the terminal stage of the condition but do not recapitulate the progressive, multifactorial nature of the human disorder. ### Genetic Models Genetic models carry engineered mutations that reproduce a known disease-associated genotype. These are particularly informative for monogenic disorders. The Mecp2-null mouse, which models Rett syndrome, has been used to study the neurological consequences of MECP2 deletion. Metabolomic analysis of brain extracts from these animals has revealed altered brain growth, perturbed osmoregulation, and changes in neurotransmitter metabolism, demonstrating how genetic models can be combined with systems-level analytical methods to uncover unexpected disease mechanisms. Genetic models offer high construct validity when the mutation is orthologous to the human cause, but their phenotypes may be subtle, variable between strains, or require aging before features emerge. ### Surgical and Lesion Models Surgical models create a defined anatomical lesion through stereotaxic injection of excitotoxins, mechanical transection, or targeted ablation. These preparations allow precise control over the location and extent of damage, which is advantageous for studying circuit function and for testing interventions delivered to a specific region. The trade-off is that surgical models require specialised expertise, carry perioperative morbidity, and reproduce only the structural component of a disease. ### Spontaneous and Induced Models Some species develop neurological conditions spontaneously, providing models with natural disease onset and progression. Induced models, such as those created by kindling stimulation in epilepsy research, generate a disease-like state through repeated experimental manipulation. The choice between spontaneous and induced models depends on the availability of affected animals, the predictability of disease onset, and the degree of experimental control required. ## Species Selection Rodents remain the most common choice for neurological research because of their short generation times, well-characterized genetics, and the availability of species-specific reagents. Their lissencephalic brains and differences in immune and metabolic physiology, however, limit their utility for certain questions. Large animal models, including pigs and sheep, offer gyrencephalic brains and closer physiological resemblance to humans, but they are more expensive, require more substantial housing, and have fewer genetic tools available. Non-human primates provide the closest phylogenetic match for complex cognitive and motor functions, yet their use raises significant welfare and ethical considerations and is restricted in many jurisdictions. The choice of species should be driven by the biological question instead of by convention. A study of hippocampal seizure detection and closed-loop neuromodulation, for example, may benefit from a large animal model because the implanted device and recording electrodes are physically comparable to human systems. Stanslaski and colleagues validated a fully implantable closed-loop neuromodulation device in a large animal model of epilepsy, demonstrating that the system could measure and detect seizure activity in the hippocampus during and after stimulation. This work illustrates how the technical demands of a study can dictate the species choice. ## Welfare and Regulatory Constraints Animal welfare oversight is an integral component of model selection, not an external restriction. The Guide for the Care and Use of Laboratory Animals establishes the institutional framework for housing, veterinary care, and experimental procedures, and researchers must design studies that comply with these standards. The NC3Rs provides practical guidance on replacement, reduction, and refinement, and these principles should inform model choice from the outset. A model that produces severe, unrelieved suffering may be scientifically informative but ethically and legally untenable, and the researcher should consider whether a less invasive preparation could answer the same question. Severity assessment is a formal requirement in many jurisdictions. The researcher must classify the expected severity of each procedure and justify that classification to the institutional animal care and use committee. Neurological procedures frequently involve stereotaxic surgery, repeated behavioral testing, and the possibility of seizures or motor impairment, all of which contribute to the severity profile. Refinement strategies, such as improved anesthesia protocols, postoperative analgesia, and early humane endpoints, should be incorporated into the study design before the protocol is submitted for approval. ## A Decision Framework for Model Selection The selection process begins with a precise definition of the research question. A study investigating the molecular pathogenesis of protein misfolding requires a different model than one screening anticonvulsant efficacy or characterizing non-motor features of a degenerative disease. The first decision point is whether the study aims to model aetiology, pathology, pathogenesis, or symptomatology. Each target maps to distinct model categories with different strengths and limitations. For Parkinson's disease research, the available models span pharmacological, toxin, genetic, and α-synuclein-based approaches. A guide to selecting among these models emphasizes that no single preparation replicates the full disease spectrum, and the investigator must prioritize which features are essential for the study question. Toxin models such as 6-hydroxydopamine and MPTP reproduce nigrostriatal degeneration and motor deficits but do not generate authentic α-synuclein pathology. Genetic models recapitulate aspects of familial disease but often lack the progressive neurodegeneration seen in sporadic cases. The choice hinges on whether the study requires fidelity to a specific pathogenic mechanism or to a behavioral phenotype. Epilepsy research presents a parallel challenge. A phenotype-oriented review of epilepsy models notes that the kindling model, historically the most widely used preparation, permits study of epileptogenesis and drug screening, but newer models have been developed to mimic specific seizure types. The absence of clear merits and demerits across these models means selection must be driven by the seizure phenotype under investigation. A model of acute seizure induction serves different purposes than a model of spontaneous recurrent seizures or one that reproduces the comorbidities of chronic epilepsy. ### Matching Model to Disease Stage Neurological diseases progress through stages, and the model must reflect the stage of interest. For a degenerative condition, early-stage models may emphasize biomarker discovery and mechanistic pathways, while late-stage models focus on behavioral deficits and therapeutic rescue. The Mecp2-null mouse model of Rett syndrome illustrates this principle. Metabolomic analysis of brain extracts from these animals revealed evidence of astrocyte involvement, reduced choline phospholipid turnover, and perturbed osmoregulation, findings that link genotype to phenotype through intermediate metabolic pathways. A study targeting early metabolic changes would require a different time point and analytical approach than one characterizing end-stage neuropathology. The stage of disease also determines which outcome measures are feasible. Acute toxin models produce rapid, predictable lesions suitable for testing neuroprotective interventions. Chronic genetic models may require months before a behavioral phenotype emerges, making them suitable for studying disease progression but resource-intensive for screening studies. The time course of the model must align with the funding period, institutional animal holding limits, and the expected trajectory of the disease process. ## Practical Assessment Sequence Once a candidate model is identified, a structured assessment sequence determines whether it is fit for purpose. The sequence proceeds through five checkpoints. First, confirm the construct validity of the model. Does the model reproduce the fundamental pathology or mechanism of the human condition? For a genetic model, this requires verification that the targeted mutation produces the expected molecular consequence. For a toxin model, it requires confirmation that the lesion develops in the intended anatomical region with the expected time course. Second, assess face validity. Does the model display the clinical features of interest? This assessment should be based on objective, quantifiable measures instead of subjective observation. Motor function can be assessed with open field activity, rotarod performance, or gait analysis. Seizure models require video-EEG monitoring to confirm electrographic correlates of behavioral events. Non-motor features, such as anxiety-like behavior or cognitive impairment, require validated behavioral paradigms appropriate to the species. Third, evaluate predictive validity. Will the model respond to therapeutic interventions in a way that predicts clinical efficacy? This is particularly important for drug screening studies. A model that fails to respond to established treatments has limited value for identifying new therapies, regardless of how well it reproduces other aspects of the disease. Fourth, consider the practical feasibility of the model in the local facility. This includes the availability of the species, the technical expertise required for model induction or breeding, the cost of maintenance, and the capacity for the required outcome measures. A model that requires sophisticated surgical expertise or specialised imaging equipment may be impractical in a facility lacking these resources. Fifth, review the welfare implications of the model. The severity of the procedure, the expected disease course, and the humane endpoints must be defined prospectively. Institutional animal care and use programs provide the framework for this review, and the [National Research Council guide for laboratory animal care](https://grants.nih.gov/grants/olaw/guide-for-the-care-and-use-of-laboratory-animals.pdf) specifies the responsibilities of the institution and the investigator in this process. ### Decision Points That Change the Choice Several factors can alter the correct model choice even when the research question appears similar. The species is the most obvious variable. Rodent models dominate neurological research because of their genetic tractability, short generation times, and lower housing costs. However, species with more complex nervous systems may be required for certain questions. Large animal models offer advantages for studying gyrencephalic brain structure, testing implantable devices, and evaluating behavioral phenotypes that are poorly represented in rodents. A fully implantable closed-loop neuromodulation device was validated in a large animal model of epilepsy, where the device measured and detected seizure activity in the hippocampus during and after stimulation. This work demonstrates that device development and validation may require species with brain dimensions sufficient to accommodate the hardware. The production system or patient status also influences model selection. A study targeting a naturally occurring neurological disease in livestock or companion animals may use the spontaneous disease as the model, instead of inducing a condition in a laboratory species. This approach offers advantages in translational relevance but requires access to clinical populations and standardized phenotyping protocols. The [MSD Veterinary Manual](https://www.msdvetmanual.com/) provides species-specific clinical information that can inform the characterization of naturally occurring disease models. The available equipment and analytical platforms constrain the choice of outcome measures. A facility with magnetic resonance spectroscopy capacity can pursue metabolomic endpoints, as demonstrated in the Rett syndrome study. A facility limited to behavioral assessment and histology must select models with robust behavioral phenotypes and well-characterized pathological markers. The model must be chosen to match the measurement capabilities of the laboratory, not the reverse. ## Model Characterization and Baseline Data Before experimental manipulation begins, the model must be characterized under local conditions. This includes establishing baseline values for all outcome measures, defining the natural history of the model, and identifying potential confounders. For genetic models, this requires confirming the genotype and assessing the phenotype in the specific genetic background, since background strain effects can substantially alter disease expression. For induced models, this requires standardizing the induction protocol and verifying the consistency of the resulting lesion or pathology. Baseline characterization should include sex as a biological variable. Many neurological disease models show sex differences in disease severity, progression, or response to treatment. The choice of single-sex or mixed-sex cohorts must be justified by the research question and the known biology of the model. The [AVMA professional practice resources](https://www.avma.org/resources-tools) provide guidance on study design considerations that apply to animal research. ## Monitoring Parameters and Their Interpretation The monitoring plan for a neurological model must capture both the disease process and the animal's welfare status. The specific parameters depend on the model and the species, but several categories apply broadly. | Monitoring Category | Parameters | What It Detects | Frequency | |---|---|---|---| | Neurological status | Gait, posture, righting reflex, proprioception, seizure activity | Progression of motor deficits, onset of paroxysmal events, deterioration of central nervous system function | Daily to weekly depending on model | | Behavioral change | Activity level, grooming, social interaction, response to handling | Non-motor features, pain, distress, humane endpoint criteria | Daily | | Body condition | Body weight, body condition score, food and water intake | Systemic effects of the disease, intercurrent illness, welfare compromise | Twice weekly minimum | | Physiological measures | Temperature, heart rate, respiratory rate | Autonomic dysfunction, systemic inflammation, decompensation | As indicated by clinical status | | Electrographic data | EEG, local field potentials | Seizure activity, abnormal network oscillations, response to neuromodulation | Continuous or scheduled recording | | Biochemical markers | Blood or cerebrospinal fluid biomarkers | Disease progression, target engagement, off-target effects | At defined study time points | The interpretation of monitoring data must be pre-specified. Thresholds for intervention, humane euthanasia, or removal from the study should be defined before the experiment begins and approved through the institutional animal care and use process. The [NC3Rs resources on the 3Rs](https://www.nc3rs.org.uk/) provide practical guidance on refining procedures and establishing welfare assessment protocols that minimize animal suffering while preserving scientific validity. ## Documentation and Data Integrity The documentation of model selection, characterization, and experimental outcomes must be sufficiently detailed to allow replication. This includes the source of the animals, genetic background, housing conditions, induction protocols, anesthetic and analgesic regimens, and the timing of all assessments. The [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) establish expectations for animal health and welfare documentation that apply to international collaborative research. The recording of neurological outcome measures requires standardized protocols. Video recording of behavior allows blinded scoring and inter-rater reliability assessment. Electrographic data require consistent acquisition parameters and clearly defined event detection criteria. Histological endpoints require systematic sampling and quantitative analysis with pre-specified criteria for positive findings. Data management should include version control, secure storage, and a clear chain of custody for all primary records. The study protocol, amendments, and deviations must be documented contemporaneously. This documentation supports both scientific integrity and regulatory compliance, and it enables the critical evaluation of the model's performance that informs future model selection decisions. ## Recognized Complications and Failure Modes Model failure rarely announces itself. It emerges as drift in baseline measurements, unexpected mortality, or a treatment effect that cannot be reproduced. The most common failure mode is phenotypic drift in genetic lines. Breeding colonies accumulate modifier mutations, and the expression of a targeted deletion changes over generations. Detect this by maintaining cryopreserved stocks of early-passage embryos and by rederiving colonies at defined intervals. Compare behavioral and biochemical phenotypes against the original characterization data at each rederivation. A second failure mode is the unintended lesion. Stereotactic coordinates are derived from atlas brains, and strain-specific skull geometry shifts target accuracy. A guide for laboratory animal care and use emphasizes that surgical competence and aseptic technique are institutional responsibilities, not investigator preferences. Verify placement histologically in every animal, not in a pilot cohort. If a toxin model depends on injection site accuracy, a 0.5 mm error can convert a partial lesion into a complete one or spare the target entirely. The third failure mode is pharmacological tolerance or compensatory adaptation. Toxin models such as 6-hydroxydopamine produce stable lesions, but the remaining dopaminergic neurons upregulate synthesis and release. A behavioral readout may normalize while the lesion persists. Conversely, kindling models show progressive intensification of seizures, so the model phenotype changes with each stimulation. The phenotype-oriented review of epilepsy models notes that different models mimic different seizure types without clear merits and demerits, which means the investigator must define which seizure feature is the experimental endpoint before choosing the model. ## Common Errors in Model Selection Less experienced investigators often choose a model because it is familiar instead of because it matches the question. A student studying non-motor symptoms of Parkinson's disease who selects a toxin model that reproduces only motor deficits has made a category error. The guide to selecting Parkinson's disease models organizes available models by which features each replicates, allowing the investigator to match aetiology, pathology, and symptom profile directly to the research question. A second error is treating the species as interchangeable. The genetic background of a mouse strain alters seizure threshold, inflammatory response, and drug metabolism. The same mutation on a C57BL/6 versus a 129 background can produce different phenotypes. Corrective action is to consult strain-specific literature before committing to a breeding strategy. A third error is underpowered baseline characterization. Investigators collect baseline data on a small cohort and assume it applies to the entire colony. The corrective action is to establish normative ranges for each behavioral assay, imaging modality, and biochemical marker in the specific strain, sex, and age range to be used. The metabolomic study of Mecp2-null mice demonstrates that even a small number of animals can reveal meaningful metabolic alterations when the analytical method is sufficiently sensitive, but the same study also shows that interpretation depends on comparison with matched wild-type controls. ## Limitations of the Current Evidence The evidence base for animal model validity is uneven. Some disorders, such as Parkinson's disease, have a wide range of models with documented strengths and limitations. Others rely on a single genetic line or a single toxin protocol. Expert opinion differs on whether genetic models that reproduce the molecular pathology of a disease are inherently more valuable than pharmacological models that reproduce the behavioral phenotype. The choice depends on whether the research question concerns mechanism or treatment efficacy. A further limitation is the translational gap. A model that responds to an intervention does not guarantee that the intervention will work in human patients. The closed-loop neuromodulation device validated in a large animal model of epilepsy demonstrates that device performance can be assessed chronically in vivo, but the same study notes that stimulation interferes with neural sensing, a problem that required hardware and algorithmic solutions. These technical constraints are rarely described in the primary literature and are often learned only through direct experience. ## When to Refer or Consult Referral is warranted when the model requires expertise outside the core laboratory. Stereotactic surgery in neonatal rodents, chronic electrode implantation, and behavioral phenotyping each demand specialised training. A laboratory animal veterinarian should be consulted when a procedure causes unexpected morbidity or when analgesic or anesthetic protocols need adjustment for a specific strain or age group. Regulatory reporting obligations vary by jurisdiction. The [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) define reporting requirements for notifiable diseases, and the [AVMA practice resources](https://www.avma.org/resources-tools) provide guidance on professional obligations. Institutional animal care and use committees must be notified of unanticipated adverse events, protocol deviations, and any mortality that exceeds the approved threshold. The [NC3Rs resources on replacement, reduction and refinement](https://www.nc3rs.org.uk/) offer practical guidance on refining procedures to prevent adverse events before they occur. ## Troubleshooting Table | Observation | Likely Cause | Discriminating Check | |---|---|---| | Progressive decline in baseline behavior | Phenotypic drift or intercurrent disease | Compare against cryopreserved stock, request health surveillance records | | Lesion size varies between cohorts | Stereotactic coordinate error or toxin batch variation | Histological verification, assay toxin activity before each use | | Seizure threshold falls over time | Kindling progression or electrode migration | Electrophysiological confirmation of electrode position, review stimulation parameters | | Treatment effect in one sex only | Sex-specific metabolism or hormonal confound | Repeat with balanced sex groups, consult strain-specific literature | | Mortality exceeds approved threshold | Anesthetic complication or post-operative infection | Review anesthetic records, request veterinary pathology consultation | | Behavioral readout normalizes despite persistent lesion | Compensatory adaptation | Terminal biochemical or histological confirmation of lesion extent | ## Frequently Asked Questions ### How Do I Choose a Model When Funding Limits Exclude the Ideal Option? Prioritize the model feature most central to your research question, then accept the best available compromise. If a non-human primate model of Parkinson's disease is unaffordable, a toxin-based rodent model may still reproduce the dopaminergic degeneration relevant to your hypothesis, even if it lacks the full spectrum of non-motor features. Document the compromise explicitly in the protocol and in any resulting publications. The [NC3Rs guidance on replacement, refinement and reduction](https://www.nc3rs.org.uk/) offers practical strategies for achieving scientific objectives with smaller or less expensive models. Budget constraints do not justify poor welfare, but they do require transparent reporting of how model limitations affect interpretation. ### What Should I Do When the Required Equipment, Such as EEG or Imaging, Is Unavailable? Adapt the outcome measures instead of abandoning the model. For epilepsy research, video-based behavioral seizure scoring can substitute for EEG when electrophysiology is unavailable, though it captures only generalized motor seizures and misses non-convulsive events. Consider whether a collaborating institution can provide access to specialised equipment. The [phenotype-oriented review of epilepsy models](https://pubmed.ncbi.nlm.nih.gov/35111370/) describes how different seizure types require different detection methods, so match your recording capability to the seizure phenotype you intend to study. If the equipment gap compromises your ability to answer the primary question, redesign the study around a model whose key features are measurable with available resources. ### How Does Model Selection Differ for Large Animal Species Compared With Rodents? Large animal models offer gyrencephalic brains, closer anatomical similarity to humans, and the ability to implant chronic recording devices, as demonstrated by a fully implantable closed-loop neuromodulation system validated in a large animal epilepsy model over one year. However, they require substantially greater housing space, longer developmental timelines, higher costs, and more complex welfare oversight. Rodent models provide genetic tractability, larger cohort sizes, and faster experimental cycles. The [Guide for the Care and Use of Laboratory Animals](https://grants.nih.gov/grants/olaw/guide-for-the-care-and-use-of-laboratory-animals.pdf) specifies species-appropriate housing and environmental enrichment that must be considered before committing to a large animal program. Choose the smallest species that can answer the scientific question, and escalate to larger species only when the neuroanatomical or physiological feature under study is absent in rodents. ### What Records Must I Keep for Neurological Model Studies? Maintain three parallel record streams: animal-level clinical data, experimental data, and protocol compliance documentation. Animal-level records should include daily neurological examination findings, body weight, food and water intake, seizure frequency if applicable, and any adverse events. Experimental records must capture baseline measurements before intervention, all monitoring parameters with timestamps, and raw data files with version control. Protocol compliance records should document anesthetic events, surgical procedures, and humane endpoint assessments. The [AVMA practice resources](https://www.avma.org/resources-tools) provide guidance on veterinary medical records that applies to research animals. Institutional animal care and use committees will audit these records, and incomplete documentation is a common reason for protocol suspension. ### How Do I Explain Model Limitations to a Client or Supervising Clinician? Frame the explanation around what the model can and cannot answer. State the specific research question first, then describe how the chosen model addresses that question, and finally name the features the model does not reproduce. For example, a Mecp2-null mouse models Rett syndrome at the genetic and metabolic level, but it does not recapitulate the full clinical trajectory seen in affected girls. The [metabolic fingerprinting study of Mecp2-null mice](https://pubmed.ncbi.nlm.nih.gov/17237885/) illustrates how a model can reveal mechanistic pathways even when the behavioral phenotype is incomplete. Emphasize that translational value depends on matching the model to the mechanism under investigation, not on reproducing every clinical feature. Offer to share the primary literature supporting the model choice. ### When Should I Reconsider or Abandon a Model Mid-Study? Abandon a model when the phenotype fails to develop within the expected timeframe, when the severity of the phenotype exceeds humane limits, or when baseline variability is so high that the required sample size becomes unattainable. Establish a priori criteria for these decisions during protocol review. Unexpected mortality, progressive weight loss, or self-trauma in a neurological model warrants immediate veterinary assessment and possible euthanasia. The [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) emphasize that welfare considerations can override experimental objectives. If a model fails, report the negative result instead of repeating the experiment with modified endpoints, as this contributes to the evidence base for model selection and reduces future animal use. ## Related Clinical & Scientific Guides * [Refining IACUC Protocols to Minimize Animal Pain and Distress](/knowledge/veterinary-medicine/laboratory-animal-science/refining-iacuc-protocols-minimize-animal-pain-distress) * [Health Monitoring Programs for Laboratory Animal Facilities](/knowledge/veterinary-medicine/laboratory-animal-science/health-monitoring-programs-for-laboratory-animal-facilities) * [Anesthetic Risk Assessment in Laboratory Animals: Preoperative Evaluation](/knowledge/veterinary-medicine/laboratory-animal-science/anesthetic-risk-assessment-in-laboratory-animals-preoperative-evaluation) ## References and Further Reading - [Animal models of Parkinson's disease: a guide to selecting the optimal model for your research.](https://pubmed.ncbi.nlm.nih.gov/34956652/). 2021. - [Animal Models of Epilepsy: A Phenotype-oriented Review.](https://pubmed.ncbi.nlm.nih.gov/35111370/). 2022. - [Metabolic fingerprints of altered brain growth, osmoregulation and neurotransmission in a Rett syndrome model.](https://pubmed.ncbi.nlm.nih.gov/17237885/). 2007. - [Neuro-QOL: quality of life item banks for adults with neurological disorders: item development and calibrations based upon clinical and general population testing.](https://pubmed.ncbi.nlm.nih.gov/21874314/). 2012. - [Design and validation of a fully implantable, chronic, closed-loop neuromodulation device with concurrent sensing and stimulation.](https://pubmed.ncbi.nlm.nih.gov/22275720/). 2012. - [Targeting RNA with small molecules: from fundamental principles towards the clinic.](https://pubmed.ncbi.nlm.nih.gov/33458725/). 2021. - [Guide for the Care and Use of Laboratory Animals, 8th Edition](https://grants.nih.gov/grants/olaw/guide-for-the-care-and-use-of-laboratory-animals.pdf). National Academies Press, 2011. - [NC3Rs Resources on Replacement, Reduction and Refinement](https://www.nc3rs.org.uk/). NC3Rs. - [MSD Veterinary Manual, Professional Edition](https://www.msdvetmanual.com/). MSD Veterinary Manual. ## Related Articles - [Selecting Animal Models for Infectious Disease Research](/knowledge/veterinary-medicine/laboratory-animal-science/selecting-animal-models-for-infectious-disease-research) - [Selecting Animal Models for Neurological Research](/knowledge/veterinary-medicine/laboratory-animal-science/selecting-animal-models-neurological-research) - [Scoring Severity of Procedures in Animal Research Protocols](/knowledge/veterinary-medicine/laboratory-animal-science/scoring-severity-procedures-animal-research-protocols) - [Selecting Appropriate Animal Models for Pain Research](/knowledge/veterinary-medicine/laboratory-animal-science/selecting-appropriate-animal-models-for-pain-research) - [Applying the 3Rs in Veterinary Research: Practical Examples](/knowledge/veterinary-medicine/laboratory-animal-science/applying-the-3rs-in-veterinary-research-practical-examples) > 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.