# Cohort Studies in Veterinary Research: Design and Interpretation


## Key Takeaways

- Cohort studies establish temporality by classifying exposure status before outcome ascertainment, allowing for direct calculation of incidence and relative risk, crucial for understanding disease progression and prognostic factors in veterinary populations.
- Prospective cohort studies offer strong control over data collection, enabling precise measurement of exposure and outcomes, but are limited by time and potential for attrition, particularly in companion animal populations where owner mobility can lead to loss to follow-up.
- Retrospective cohort studies leverage existing records to efficiently study long-latency outcomes or rare exposures, but are constrained by the completeness, accuracy, and potential for differential ascertainment inherent in historical data.
- Ambidirectional cohort designs integrate retrospective and prospective data collection, proving valuable when past exposures require ongoing outcome observation, as demonstrated in studies of long-term survival in hyperthyroid cats treated with radioiodine.
- Key threats to validity in cohort studies include selection bias at enrollment, differential loss to follow-up, information bias from measurement error (e.g., misclassification of exposure or outcome), and confounding, which must be addressed through careful design and analysis, often guided by STROBE-Vet reporting standards.
- Common veterinary applications include assessing survival after treatment, identifying incidence of diseases in exposed versus unexposed groups (e.g., management practices and respiratory disease in feedlot cattle), and evaluating prognostic factors for conditions like neoplasia or chronic degenerative diseases.

---

Cohort studies occupy a central position in veterinary epidemiology because they allow direct observation of disease occurrence, progression, and outcome in defined animal populations over time. This article explains the design logic, conduct, and interpretation of cohort studies in veterinary settings for researchers who plan, appraise, or apply such evidence. It covers prospective, retrospective, and ambidirectional designs, addresses the selection and measurement decisions that determine study validity, and describes the analytical frameworks used to quantify risk and survival. The content applies across species and production systems, from companion animal practice to livestock populations, and assumes familiarity with basic epidemiological terminology.

The central question a cohort study answers is whether an exposure or risk factor is associated with the subsequent development of an outcome. Unlike case-control designs, which begin with diseased and non-diseased individuals and look backward, cohort studies assemble a group of animals, classify them by exposure status, and follow them forward in time to observe outcome occurrence. This temporal directionality is the defining feature and the principal source of the design's strength in establishing causal sequence. A veterinary researcher might use a cohort design to ask whether early-life neutering alters the incidence of neoplasia, whether a particular management practice predicts respiratory disease in feedlot cattle, or whether clinicopathological variables at diagnosis predict survival in hyperthyroid cats treated with radioiodine.

## At a Glance

| Parameter | Decision or fact |
|---|---|
| Core design feature | Exposure classified before outcome ascertainment, animals followed forward in time |
| Prospective cohort | Exposure and covariates measured at enrollment, then follow-up into the future |
| Retrospective cohort | Exposure and outcome already occurred, reconstructed from records |
| Ambidirectional cohort | Combines retrospective and prospective data collection phases |
| Primary measure of association | Relative risk (cumulative incidence ratio) or hazard ratio from time-to-event analysis |
| Key validity threats | Selection bias at enrollment, loss to follow-up, information bias, confounding |
| Reporting standard | STROBE-Vet, an extension of the STROBE statement for veterinary observational studies |
| Common veterinary applications | Survival after treatment, disease incidence in exposed versus unexposed groups, prognostic factor identification |

## Definition and Conceptual Basis

A cohort is a group of animals that shares a defining characteriztic or experience within a defined time period and is observed over time. The term derives from the Latin *cohors*, a military unit, and the epidemiological usage retains the sense of a bounded group moving forward together. Cohort membership is fixed at the start of follow-up, although the exposure of interest may be measured at baseline, may change during follow-up, or may be accumulated over time as a time-varying quantity.

The fundamental logic rests on the principle that exposure status is established before the outcome occurs. This temporal separation allows the researcher to calculate the incidence of the outcome in exposed and unexposed groups and to compare those incidences directly. The comparison yields the relative risk, which estimates the strength of association between exposure and outcome. Because the design can accommodate multiple outcomes from a single exposure, cohort studies are efficient when the outcome of interest is common or when several outcomes are of interest simultaneously.

## Prospective Cohort Studies

In a prospective cohort study, the researcher identifies the study population at the present time, measures exposure and baseline covariates, and then follows animals forward in time until the outcome occurs, the study ends, or animals are lost to follow-up. This design offers the strongest control over data collection because the researcher specifies the measurement protocols, defines the exposure and outcome criteria before data collection begins, and can ensure that outcome ascertainment is blinded to exposure status where feasible.

The principal disadvantage is the time required. Outcomes with long latency periods, such as neoplasia or chronic degenerative disease, may require years of follow-up before sufficient events accumulate. Attrition is a corresponding concern, particularly in companion animal populations where owners may move, change veterinary practices, or elect euthanasia for reasons unrelated to the study outcome. The [ARRIVE guidelines for reporting animal research](https://arriveguidelines.org/) emphasize the importance of describing follow-up procedures and attrition so that readers can assess the risk of bias from incomplete observation.

Prospective designs are well suited to questions about incidence in defined populations, such as the development of behavioral or health outcomes in dogs followed from puppyhood, or the occurrence of production diseases in a monitored livestock cohort. The researcher must define the source population, the sampling frame, and the eligibility criteria before enrollment, and must decide whether the cohort is fixed, with all animals enrolled at the same time, or dynamic, with animals entering over a period.

## Retrospective Cohort Studies

A retrospective cohort study identifies a cohort from historical records, classifies exposure status from data recorded at the time, and then determines outcomes that have already occurred. The defining feature is that both exposure and outcome are in the past when the study begins. This design is efficient when the outcome has a long latency, when the exposure is rare, or when the researcher has access to high-quality records from a defined population.

The retrospective approach depends entirely on the completeness and accuracy of pre-existing data. Medical records, herd health databases, and registries vary in the consistency of their data collection, and variables that were not recorded cannot be recovered. The researcher must also confront the possibility that animals with certain exposures were monitored more intensively, creating differential outcome ascertainment. The [EQUATOR Network reporting guidelines library](https://www.equator-network.org/) includes the STROBE statement and its veterinary extension, which specify the information needed to evaluate the adequacy of record-based exposure and outcome measurement.

A retrospective cohort design was used to examine whether childhood exposure to pet dogs or cats was associated with the later diagnosis of mental health disorders in adolescence, merging an earlier study database with electronic medical records to identify diagnoses using ICD-9 and ICD-10 codes. The investigators used proportional hazards regression to compare time to diagnosis between youth with and without pets, illustrating how retrospective cohorts can leverage existing data infrastructure to address questions that would be impractical prospectively.

## Ambidirectional Cohort Studies

Some research questions benefit from combining retrospective and prospective elements. An ambidirectional cohort study identifies the cohort and measures baseline exposure retrospectively, then follows animals forward to collect outcomes and additional covariates prospectively. This hybrid design is particularly useful when the exposure of interest occurred in the past but the outcome requires ongoing observation.

A study of 231 hyperthyroid cats treated with radioactive iodine at a veterinary teaching hospital used an ambidirectional design, collecting data at the time of diagnosis retrospectively and then following cats for a median of 25 months to determine survival. Cox proportional hazards models identified age at diagnosis and sex as predictors of survival, with increasing age and male sex associated with greater likelihood of death. The design allowed the investigators to include both historical diagnostic data and prospectively collected information on health problems that developed after treatment, demonstrating how the two data collection phases can be integrated in a single analytical model.

The choice among prospective, retrospective, and ambidirectional designs depends on the latency of the outcome, the availability of records, the resources for follow-up, and the urgency of the research question. Each design shares the same underlying logic of exposure classification preceding outcome ascertainment, but the practical conduct and the specific threats to validity differ substantially.

## Selecting a Cohort Design: A Decision Framework

The choice between prospective, retrospective, and ambidirectional designs is rarely a matter of preference. It follows from the research question, the availability and quality of existing records, the latency of the outcome, and the resources at hand. A decision table can make these trade-offs explicit.

| Decision factor | Prospective cohort preferred | Retrospective cohort preferred |
|---|---|---|
| Outcome latency | Short to moderate follow-up feasible within funding cycle | Long latency outcomes where historical exposure data exist |
| Exposure ascertainment | Direct measurement, standardized protocols, biological sampling needed | Exposure already documented in medical records, registries, or production databases |
| Outcome ascertainment | Active surveillance with scheduled re-examination | Passive surveillance using existing diagnostic codes, laboratory records, or slaughter data |
| Baseline data quality | High, because data collection is designed for the study | Variable, depends on record completeness and consistency |
| Loss to follow-up risk | High, especially in client-owned animals over years | Lower, because outcomes may already have occurred |
| Cost and timeline | Higher cost, longer timeline | Lower cost, faster results |
| Causal inference strength | Stronger, because temporality is clear and measurement is controlled | Weaker, because exposure and outcome may both be affected by record-keeping patterns |
| Rare outcomes | Inefficient unless outcome is common or cohort is very large | More efficient if historical cohorts with adequate outcome frequency can be assembled |

A practical example illustrates the trade-off. A study of survival after radioactive iodine treatment for hyperthyroid cats used an ambidirectional design, collecting historical data at diagnosis and then following cats forward in time. This allowed the investigators to model predictors of survival using both baseline variables and health problems that developed during follow-up. The design was appropriate because the exposure of interest, treatment, had already occurred, but the outcomes of interest, long-term survival and incident comorbidities, required prospective observation. See the [long-term survival study of hyperthyroid cats treated with iodine 131](https://pubmed.ncbi.nlm.nih.gov/11215911/) for the full methodology.

Species and production system change the calculus. In companion animal practice, electronic medical records are often incomplete across clinics, and client mobility creates substantial loss to follow-up. A prospective design with scheduled rechecks and owner contact protocols may be the only way to ensure outcome ascertainment. In production medicine, by contrast, herd records, slaughter data, and regulatory tracing systems may permit efficient retrospective assembly of large cohorts. However, the investigator must verify that the records capture the exposure with sufficient detail and that the population base is well defined. The [World Organization for Animal Health terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/) provide relevant frameworks for surveillance data quality in livestock and wildlife populations.

Patient status also matters. For conditions with high early mortality, such as severe trauma or peracute infectious disease, a retrospective cohort assembled from hospital admissions may miss animals that died before presentation. A prospective design with community-based recruitment may be necessary to capture the full spectrum of disease. Conversely, for chronic conditions in geriatric patients, the competing risk of death from unrelated causes complicates both designs, and the analysis must account for this through appropriate censoring.

## Assembling the Cohort and Defining Exposure

The cohort must be defined by a clear eligibility criterion that does not depend on the outcome. In veterinary studies, the source population is often a hospital population, a referral caseload, or an insured animal population. Each has selection pressures. Hospital populations over-represent severe disease and may include animals with prior treatment that modifies the exposure. Insured populations skew toward owners with financial capacity and may under-represent certain breeds or management systems. These selection effects should be described explicitly in the methods and considered in the interpretation.

Exposure definition requires the same rigour as outcome definition. A study of dietary protein intake and muscle mass loss in older people illustrates the principle: the investigators defined protein intake using a validated food frequency questionnaire, expressed it as grams per kilogram of body weight, and divided it into quartiles for analysis. This approach allowed a dose-response assessment instead of a binary exposed or unexposed comparison. Veterinary researchers should adopt similar practices, using validated instruments where they exist and reporting the distribution of exposure levels instead of collapsing continuous variables prematurely. See the [four-year cohort study of dietary protein intake and physical decline](https://pubmed.ncbi.nlm.nih.gov/24522470/) for an example of exposure categorisation and adjustment for confounders.

Exposure measurement error is a particular hazard in retrospective cohorts. Clinical records may record drug doses, but not whether the owner administered them. Production records may document vaccination dates, but not cold chain integrity or concurrent disease. Where possible, the study should include a validation substudy comparing recorded exposure against an independent source, such as owner interview, pharmacy dispensing records, or serological evidence.

## Outcome Ascertainment and Follow-Up

Outcome definitions must be specified before data analysis, with explicit diagnostic criteria. In retrospective cohorts, outcomes are often identified through diagnostic codes, which can misclassify cases if coding practices vary between clinicians or over time. The [study of pet exposure and adolescent mental health diagnoses](https://pubmed.ncbi.nlm.nih.gov/36199055/) identified outcomes using ICD-9 and ICD-10 codes from electronic medical records, a pragmatic approach that trades diagnostic precision for sample size. Veterinary researchers using diagnostic codes should audit a sample of records to confirm that the codes correspond to the clinical findings.

Follow-up protocols in prospective cohorts should specify the schedule of re-examination, the methods used to contact owners, and the procedures for animals that move or change ownership. Loss to follow-up is rarely random. Animals that die at home, are euthanised for financial reasons, or are sold may differ systematically from those that complete the study. The analysis should compare baseline characteriztics of animals with complete follow-up against those lost, and sensitivity analyzes should explore the impact of plausible missing data mechanisms.

## Bias and Confounding in Cohort Studies

Selection bias enters cohort studies when the probability of inclusion depends on both exposure and outcome. In retrospective cohorts assembled from referral populations, this can occur if referral patterns are influenced by the outcome of interest. For example, a study of postoperative complications using a referral hospital cohort may over-represent animals with complications, because primary care veterinarians refer complicated cases. The direction and magnitude of this bias should be discussed.

Information bias arises from differential measurement error. In prospective studies, outcome assessors should be blinded to exposure status where feasible. In retrospective studies, the investigator should check whether diagnostic workup intensity differed between exposure groups. An exposed group that receives more frequent monitoring will have more opportunities to have outcomes detected, a form of surveillance bias.

Confounding is addressed at the design stage through restriction or matching, and at the analysis stage through stratification or multivariable regression. The [hyperthyroid cat survival study](https://pubmed.ncbi.nlm.nih.gov/11215911/) used Cox proportional hazards models to adjust for age and sex, and the investigators reported relative risks with confidence intervals, allowing readers to judge the precision of the estimates. Veterinary cohort studies should follow this model, reporting effect measures with confidence intervals instead of relying on p-values alone.

## Reporting and Interpretation Checklist

The STROBE statement, available through the [EQUATOR Network reporting guidelines library](https://www.equator-network.org/), provides a structured checklist for reporting observational studies. Veterinary researchers should also consult the [ARRIVE guidelines for reporting animal research](https://arriveguidelines.org/), which specify the minimum information required for transparent and reproducible publications. The following checklist is adapted for veterinary cohort studies:

- State the study design and the source population with eligibility criteria.
- Describe how the cohort was assembled and the time period of enrollment.
- Define exposures and outcomes with explicit criteria, and state whether assessors were blinded.
- Report the number of animals at each stage, including exclusions and losses to follow-up.
- Provide baseline characteriztics of the cohort by exposure group.
- Specify the statistical methods, including how confounding was addressed.
- Report effect measures with confidence intervals and the number of events.
- Discuss limitations, including selection bias, information bias, and generalizability.
- State the funding source and any conflicts of interest.

The [MSD Veterinary Manual](https://www.msdvetmanual.com/) and the [American Veterinary Medical Association practice resources](https://www.avma.org/resources-tools) provide species-specific context that can help investigators anticipate practical obstacles in outcome ascertainment and follow-up, particularly in private practice settings where research infrastructure is limited.

## Recognized Complications and Failure Modes

Cohort studies fail in characteriztic patterns. The most consequential is differential loss to follow-up, where animals with a particular exposure or outcome are disproportionately withdrawn from observation. In a prospective study of hyperthyroid cats treated with iodine 131, the authors maintained a median follow-up of 25 months and used Cox proportional hazards models to identify predictors of survival, but such analyzes cannot fully repair attrition that is informative [Long-term health and predictors of survival for hyperthyroid cats treated with iodine 131](https://pubmed.ncbi.nlm.nih.gov/11215911/). Early detection requires a cumulative follow-up table at each scheduled visit, comparing the proportion of exposed and unexposed animals still under observation. When attrition exceeds 20 percent overall or differs by more than 10 percentage points between exposure groups, the study should be assessed for selection bias before analysis proceeds.

Misclassification of exposure is a second failure mode. In retrospective cohorts, exposure is reconstructed from records that were not designed for research, so a dog that received a vaccine at an outside clinic may be recorded as unvaccinated. Detection relies on a validation substudy in which a random sample of records is reabstracted by a second observer blinded to the original coding, with agreement quantified by the kappa statistic. A kappa below 0.60 should trigger revision of the abstraction instrument.

Outcome misclassification is equally damaging. When outcomes are extracted from electronic medical records using diagnostic codes, as in the retrospective cohort study of childhood pet exposure and adolescent mental illness, the sensitivity and specificity of those codes must be established against a reference standard such as full chart review [Impact of pet dog or cat exposure during childhood on mental illness during adolescence: a cohort study](https://pubmed.ncbi.nlm.nih.gov/36199055/). Differential misclassification, where outcome detection differs by exposure status, is particularly dangerous because it can create or mask associations.

## Common Errors and Corrective Actions

Less experienced investigators frequently confuse the temporal sequence in retrospective cohorts. They identify exposed and unexposed animals from records but fail to verify that exposure preceded outcome for every subject. The corrective action is a dated eligibility checklist applied to each animal before enrollment, with exposure date and outcome date confirmed against source documents.

A second common error is treating a retrospective cohort as a case-control study by sampling on outcome. If the investigator selects all animals with the outcome and a convenience sample of those without it, the result is no longer a cohort and the risk estimates are not interpretable as incidence. The corrective action is to define the cohort as a complete enumeration of an eligible population, or a random sample of that population, before any outcome is considered.

A third error is overadjustment in multivariable models. Including variables that lie on the causal pathway between exposure and outcome, such as adjusting for body condition score when studying the effect of a diet on weight gain, attenuates or eliminates the association of interest. The corrective action is to construct a directed acyclic graph before analysis and to justify each covariate as a confounder, mediator, or collider.

## Limitations of Current Evidence and Areas of Disagreement

The veterinary cohort literature is thinner than its human counterpart. Many published studies are small, single-center, and short in duration. The retrospective cross-sectional cohort design used in the Dogs Trust National Dog Survey, which drew on 354,224 classified expectation responses, illustrates both the reach and the limits of owner-reported data: the sample is large, but exposure and outcome are measured simultaneously and recall is subject to error [Owner expectations and surprises of dog ownership experiences in the United Kingdom](https://pubmed.ncbi.nlm.nih.gov/38384957/).

Expert opinion still differs on the minimum acceptable follow-up duration for chronic disease outcomes, on whether owner-reported outcomes can substitute for clinical measurements, and on the role of propensity score methods in observational veterinary studies. Reporting standards such as the ARRIVE guidelines and the EQUATOR Network's STROBE checklist provide a common framework, but adherence in the veterinary literature remains uneven [ARRIVE Guidelines 2.0 for Reporting Animal Research](https://arriveguidelines.org/), [EQUATOR Network Reporting Guidelines](https://www.equator-network.org/). Where the evidence base is contested, the responsible approach is to state the uncertainty explicitly and to report sensitivity analyzes.

## Referral, Consultation, and Regulatory Reporting

Most cohort studies do not require referral, but several circumstances warrant specialist input. A veterinary epidemiologist or biostatistician should be consulted before the protocol is finalised when the study involves complex sampling, clustered data, or time-to-event analysis. A veterinary pathologist should review outcome definitions when the endpoint is histopathological. Laboratory involvement is required when diagnostic tests are used to define exposure or outcome, and assay validation data should be reported.

Regulatory reporting obligations arise when a study identifies a notifiable disease, a suspected adverse drug event, or an emerging welfare concern. Requirements vary by jurisdiction and species, and investigators should confirm their obligations with the relevant authority before commencing data collection. The World Organization for Animal Health maintains international standards for disease surveillance and reporting that apply to WOAH-listed diseases [WOAH terrestrial animal health standards](https://www.woah.org/en/what-we-do/standards/codes-and-manuals/terrestrial-code-online-access/). Institutional animal care and use approval must be obtained before any prospective data collection begins, and data protection obligations apply to owner-identifiable information.

| Observation | Likely cause | Discriminating check |
| --- | --- | --- |
| Attrition exceeds 20 percent or differs between exposure groups | Differential loss to follow-up | Compare baseline characteriztics of retained and lost animals |
| Exposure prevalence implausibly high or low | Misclassification from incomplete records | Validation substudy with blinded reabstraction |
| Association appears only after adjustment | Overadjustment or collider bias | Directed acyclic graph review of the adjustment set |
| Outcome codes disagree with chart review | Diagnostic code misclassification | Calculate sensitivity and specificity against full records |
| Risk estimate changes when the cohort definition is altered | Selection bias at enrollment | Repeat analysis with a restricted or expanded eligibility definition |

## Frequently Asked Questions

### How Many Animals Do I Need to Follow in a Cohort Study?

Sample size depends on the expected event rate, the magnitude of effect you wish to detect, and the precision required. For rare outcomes, you may need thousands of animals, for common outcomes in production settings, hundreds may suffice. Run a formal sample size calculation during protocol development, specifying the assumed baseline risk and a clinically meaningful effect size. Consult the [ARRIVE guidelines for reporting animal research](https://arriveguidelines.org/) when planning, as they emphasize transparent justification of sample sizes. If funding limits the achievable sample, consider extending follow-up duration or using an ambidirectional design, as illustrated by a study of hyperthyroid cats that combined retrospective and prospective data collection to maximize the use of available cases ([long-term survival predictors in radioiodine-treated cats](https://pubmed.ncbi.nlm.nih.gov/11215911/)).

### Can I Use Electronic Medical Records to Build a Retrospective Cohort?

Yes, but the quality of your cohort depends entirely on the completeness and consistency of the records. Define exposure and outcome variables before data extraction, and verify that the relevant fields were recorded systematically across all participating clinics. Missing data are common in practice records, so plan a priori how you will handle incomplete entries, whether by exclusion, imputation, or sensitivity analysis. A retrospective cohort built from electronic records has been used successfully to examine associations between childhood pet ownership and later mental health diagnoses, merging survey data with diagnostic codes ([pet exposure and adolescent mental health cohort study](https://pubmed.ncbi.nlm.nih.gov/36199055/)). Validate a random sample of extracted records against the original files to confirm accuracy.

### What Do I Do When Loss to Follow-Up Is High?

First, quantify the problem precisely. Report the number and proportion of animals lost, compare their baseline characteriztics with those retained, and state whether losses differed by exposure group. If losses are differential, selection bias is likely and your effect estimates may be distorted. Perform a sensitivity analysis assuming best-case and worst-case outcomes for the lost animals to see whether your conclusions change. Consider whether the outcome could be ascertained through alternative sources, such as owner telephone interviews, referring veterinarians, or national databases. The [EQUATOR Network reporting guidelines](https://www.equator-network.org/) provide structured frameworks for describing participant flow and missing data transparently.

### How Do I Handle Competing Risks in Production Animal Cohorts?

Animals that die, are culled, or are sold before the outcome occurs are no longer at risk. Treating these events as ordinary censoring can bias estimates if the competing event is related to both exposure and outcome. Use competing risk analysis, such as the Fine and Gray subdistribution hazard model, when the outcome of interest is precluded by death or culling. For example, in a dairy cohort studying clinical disease, a cow sold for slaughter cannot later develop the disease, and the reason for sale may relate to the exposure under study. Report both cause-specific hazards and subdistribution hazards when the research question concerns prognosis. Species-specific production realities should shape the analytical plan from the outset.

### What Are the Minimum Data I Must Record at Baseline?

Record exposure status, all known confounders, and the date of entry into the cohort. For animal studies, include species, breed, age, sex, neuter status, body condition score, and relevant management factors such as housing, diet, and vaccination history. Define the start of follow-up precisely, whether that is the date of exposure, the date of diagnosis, or a calendar date. Document the source population and the method of recruitment so readers can judge generalizability. The [MSD Veterinary Manual](https://www.msdvetmanual.com/) offers species-specific guidance on which baseline variables are clinically meaningful for common conditions. A standardized case report form, piloted before the study begins, reduces between-observer variation.

### How Should I Explain Cohort Study Findings to a Client or Herd Owner?

Translate relative measures into absolute terms. A hazard ratio of 2.0 sounds alarming, but if the baseline risk is 1 in 1,000, the absolute increase is small. State the number of animals that would need to be exposed for one additional outcome to occur, if the data support such a calculation. Distinguish association from causation explicitly, and acknowledge that unmeasured factors may explain the result. For herd-level advice, frame recommendations in terms of management changes that are feasible within the owner's production system. Professional resources from the [American Veterinary Medical Association](https://www.avma.org/resources-tools) can help you communicate risk concepts clearly to lay audiences while maintaining scientific accuracy.

## 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

- [Owner expectations and surprises of dog ownership experiences in the United Kingdom.](https://pubmed.ncbi.nlm.nih.gov/38384957/). 2024.
- [Impact of pet dog or cat exposure during childhood on mental illness during adolescence: a cohort study.](https://pubmed.ncbi.nlm.nih.gov/36199055/). 2022.
- [Long-term health and predictors of survival for hyperthyroid cats treated with iodine 131.](https://pubmed.ncbi.nlm.nih.gov/11215911/). 2001.
- [Dietary flavonoid intake and colorectal cancer risk: evidence from human population studies.](https://pubmed.ncbi.nlm.nih.gov/23613386/). 2013.
- [Associations of dietary protein intake on subsequent decline in muscle mass and physical functions over four years in ambulant older Chinese people.](https://pubmed.ncbi.nlm.nih.gov/24522470/). 2014.
- [A lateral approach for sinus elevation using PRGF technology.](https://pubmed.ncbi.nlm.nih.gov/19438953/). 2009.
- [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

- [Designing Questionnaire Studies for Veterinary Research](/knowledge/veterinary-medicine/veterinary-research-methods/designing-questionnaire-studies-veterinary-research)
- [Evaluating Prognostic Factor Studies in Veterinary Medicine](/knowledge/veterinary-medicine/veterinary-research-methods/evaluating-prognostic-factor-studies-veterinary-medicine)
- [How to Write a Research Protocol for Veterinary Studies](/knowledge/veterinary-medicine/veterinary-research-methods/write-research-protocol-veterinary-studies)
- [Reporting Qualitative Research in Veterinary Science](/knowledge/veterinary-medicine/veterinary-research-methods/reporting-qualitative-research-veterinary-science)
- [Designing and Analyzing Factorial Trials in Veterinary Research](/knowledge/veterinary-medicine/veterinary-research-methods/designing-analyzing-factorial-trials-veterinary-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.