Conducting Pharmacovigilance Studies in Veterinary Medicine

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

Key Takeaways

• Veterinary pharmacovigilance relies on post-authorization surveillance to detect rare adverse events, delayed toxicities, and drug interactions that pre-authorization studies cannot fully characterize due to limited sample sizes and durations.
• Spontaneous reporting systems, while the backbone for signal detection, are inherently limited by under-reporting, reporting bias, and the absence of denominator data, necessitating disproportionality analysis (e.g., Bayesian shrinkage estimators like Multi-item Gamma Poisson Shrinker) to identify potential signals.
• Confirmatory pharmacoepidemiological studies, such as cohort or case-control designs utilizing electronic health records, insurance claims, or prescription databases, are crucial for estimating incidence rates and establishing causality by providing denominator data.
• Causality assessment in veterinary pharmacovigilance integrates temporal plausibility, dechallenge/rechallenge (where ethical), biological plausibility, and consideration of species-specific metabolism to classify events as certain, probable, possible, unlikely, or unassessable.
• Distinct considerations for food-producing species include potential impacts on human food safety, residue concerns, and withdrawal periods, requiring specialized risk assessments and surveillance methods beyond those for companion animals.
• Pharmacovigilance study protocols must pre-specify research questions, target populations, observation windows, data sources, and analytical methods to mitigate post hoc bias and ensure rigorous interpretation of findings, especially when dealing with complex factors like polypharmacy or concurrent disease.

Pharmacovigilance in veterinary medicine is the science and practice of detecting, assessing, understanding, and preventing adverse effects and other drug-related problems after a veterinary medicinal product has been authorised for use. This article provides a methodological framework for veterinary researchers who design, conduct, or interpret pharmacovigilance studies across species. It covers the conceptual foundations of post-marketing drug safety surveillance, the structure and limitations of available data sources, signal detection methods, and the design of confirmatory pharmacoepidemiological studies. The scope excludes clinical trial safety reporting, which follows distinct regulatory and statistical conventions. The intended reader is a veterinary researcher with working knowledge of clinical pharmacology and epidemiological methods who requires a practical, source-anchored guide to studying drug safety in real-world animal populations.

The central scientific problem in veterinary pharmacovigilance is that pre-authorisation studies, however well conducted, cannot fully characterize the safety profile of a drug. Rare adverse events, delayed toxicities, effects in unstudied subpopulations, and interactions with concurrent disease or polypharmacy may only emerge once a product is used broadly in the field. The discipline therefore depends on systematic collection of spontaneous reports, structured interrogation of large databases, and formal epidemiological testing of suspected associations. Understanding the strengths and weaknesses of each approach is prerequisite to producing evidence that supports regulatory decisions, label changes, or product withdrawal.

At a Glance

The Scientific Basis of Post-Marketing Safety Surveillance

Post-marketing surveillance exists because the pre-authorisation evidence base is structurally incomplete. Clinical trials in veterinary medicine typically enrol hundreds to low thousands of animals, are powered to detect common adverse events, and follow subjects for limited durations. Rare events occurring in fewer than one per several thousand treated animals will almost certainly escape detection. The historical record in human pharmacovigilance demonstrates this pattern: several drugs withdrawn for hepatotoxicity between 1997 and 2016 showed liver test abnormalities during trials, yet the serious and irreversible liver injury that prompted withdrawal was not predicted by animal models or by the trial data available at the time of approval, as documented in a review of safety data and withdrawal of hepatotoxic drugs. Veterinary medicine faces the same fundamental limitation, compounded by greater species diversity, wider variation in body size and metabolism, and less comprehensive pre-authorisation testing in each target species.

The organizational history of pharmacovigilance is instructive. Modern drug safety monitoring was organized in the early 1950s in response to blood dyscrasias associated with chloramphenicol, and the core functions of case management, signal management, and benefit-risk management developed over subsequent decades, as described in an overview of pharmacovigilance practice. Veterinary pharmacovigilance systems have followed a parallel trajectory, adapting these functions to the realities of animal populations, multiple species, and the economic and welfare contexts of veterinary practice.

The Pharmacovigilance Cycle

The discipline operates as a continuous cycle with four linked phases. Case management begins with the receipt, triage, and quality assessment of individual adverse event reports. Signal management involves the systematic detection of new or changing safety issues from accumulated reports. Signal evaluation determines whether a detected association is causal, spurious, or confounded. Benefit-risk management translates validated signals into regulatory action, label changes, or clinical guidance. Each phase has distinct methodological requirements, and the boundaries between them are often blurred in practice.

Spontaneous Reporting Systems

Spontaneous reporting remains the backbone of veterinary pharmacovigilance. Practitioners, owners, and marketing authorisation holders submit reports of suspected adverse reactions to national authorities or directly to the product sponsor. The system is passive: reports arrive without active case finding, and the absence of a report does not indicate the absence of an event. The principal analytical challenge is the lack of a denominator. The number of treated animals is generally unknown, so incidence rates cannot be calculated directly from spontaneous data. Instead, analysts use disproportionality methods that compare the proportion of reports for a specific drug-event pair against the proportion expected from the background reporting distribution.

Data Sources for Veterinary Pharmacovigilance

The choice of data source determines the questions a pharmacovigilance study can answer. Spontaneous report databases are the primary resource for signal detection, while electronic health records, insurance claims, and prescription databases support confirmatory studies with denominator data.

Spontaneous Report Databases

National and regional regulatory authorities maintain databases of adverse event reports for veterinary medicinal products. These databases typically include the suspect product, the affected species, the clinical signs, the outcome, and limited information about the animal's signalment and concurrent medications. The United States Food and Drug Administration's Adverse Event Reporting System for human drugs illustrates the structure and analytical potential of such databases: a retrospective pharmacovigilance study using the Multi-item Gamma Poisson Shrinker algorithm identified disproportionate reporting of pituitary tumors with risperidone and related antipsychotics, demonstrating how disproportionality analysis of spontaneous reports can generate signals that warrant formal evaluation. Veterinary equivalents exist in many jurisdictions, though their size, accessibility, and data quality vary considerably.

Electronic Health Records and Practice Data

Electronic health records from veterinary teaching hospitals, corporate practice groups, and practice management software provide longitudinal clinical data with defined denominators. These sources support incidence estimation, cohort studies, and case-control designs. Their limitations include non-standardized coding, incomplete capture of over-the-counter and compounded products, and selection bias toward referral populations. Linkage between prescription records and clinical outcomes is often possible but requires careful attention to data governance and record linkage methodology.

Specialised Surveillance Systems

Some jurisdictions operate targeted surveillance programs for specific products, species, or adverse event types. These may include active follow-up of treated cohorts, laboratory-based monitoring for particular toxicities, or sentinel networks of participating practices. Such systems are resource-intensive but can provide higher-quality data than passive reporting for defined safety questions.

Signal Detection Methodology

Signal detection is the systematic search for drug-event associations that exceed what would be expected by chance. The fundamental unit of analysis is the individual case report, and the fundamental limitation is that reporting is a function of both the true incidence of an event and the probability that the event is recognized, attributed to the drug, and reported.

Disproportionality Analysis

Disproportionality analysis compares the observed number of reports for a specific drug-event combination with the number expected if reports were distributed independently across all drug-event pairs in the database. The expected count is calculated from the marginal totals of the reporting contingency table. Several algorithms exist, including the reporting odds ratio, the proportional reporting ratio, and Bayesian shrinkage estimators such as the Multi-item Gamma Poisson Shrinker. The Bayesian approach computes an Empiric Bayes Geometric Mean with a confidence interval, and the lower bound of this interval is commonly used as the signal threshold. The choice of algorithm matters less than the consistency of results across methods and the clinical plausibility of the association.

Confounding and Bias in Signal Detection

Spontaneous reports are subject to multiple biases. Notoriety bias occurs when a drug receives publicity, leading to increased reporting of all its adverse events. Stimulated reporting follows label changes, Dear Doctor letters, or regulatory communications. Concurrent drug use can produce apparent associations with the wrong product. The absence of denominator data means that disproportionality measures the relative reporting frequency, not the absolute risk. A signal is a hypothesis-generating finding, and it must be interpreted with the same rigour applied to any observational association before it is used to support regulatory action.

Establishing the Study Protocol

A pharmacovigilance study begins with a written protocol that fixes the research question, the target population, the drug or drug class under investigation, and the adverse events of interest. The protocol must define the observation window, the data sources, and the analytical methods before any data are extracted. Pre-specification reduces the risk of post hoc analyzes that produce unstable findings.

The research question should be structured with the same care used in clinical epidemiology. Specify the species, breed, age class, and production setting. Specify the dose range, duration of use, and concurrent medications. Specify the outcome with operational criteria, for example hepatotoxicity defined by serum alanine aminotransferase activity above a stated multiple of the reference interval with concurrent histologic or ultrasonographic confirmation. Vague outcomes generate vague signals.

The protocol must also state the unit of analysis. Adverse event reports in veterinary medicine are often counted per animal, but repeated treatments in the same animal, herd-level events in food animals, and litter-level events in companion animals each require different analytical handling. The unit chosen affects both the numerator and the denominator of every rate calculated.

Selecting the Study Design

Spontaneous reporting databases support retrospective disproportionality analyzes, which compare the proportion of reports for a specific drug-event pair against the proportion expected from all other drugs in the database. This design is efficient for hypothesis generation but cannot establish incidence. The Multi-item Gamma Poisson Shrinker algorithm, applied to the United States FDA Adverse Event Reporting System, illustrates how adjusted observed-to-expected ratios can rank drugs by signal strength for a given adverse event, as demonstrated in a pharmacovigilance study of atypical antipsychotics and pituitary tumors Szarfman et al., FDA Adverse Event Reporting System analysis.

Cohort studies are preferable when the goal is estimating risk in a defined population. Electronic health records and practice management data allow construction of exposed and unexposed cohorts with follow-up. The exposure must be verified from prescription or dispensing records instead of owner recall. The outcome must be ascertained through a standardized process that does not depend on the clinician's willingness to file a spontaneous report.

Case-control studies are efficient for rare adverse events. Cases are identified through diagnostic records, and controls are sampled from the same source population. The key decision is control selection: controls must be at risk for the event and drawn from the same clinical population as the cases. In food animal practice, controls should be matched on herd, production stage, and season.

The table below summarizes the design choices and the conditions that favour each.

Species and production system change the feasible design. Food animal pharmacovigilance often relies on slaughterhouse data, bulk tank milk testing, and herd health records because individual animal follow-up is impractical. Companion animal studies can use practice-level electronic records with owner contact for outcome verification. In both settings, the design must accommodate the data that actually exist instead of the data the investigator wishes existed.

Data Collection and Case Ascertainment

Case ascertainment requires a case definition that is applied identically to all potential cases. The definition should include clinical signs, clinicopathologic findings, and where available, histopathology or necropsy results. A two-stage process is common: a sensitive screening definition identifies candidate cases, and a specific confirmation stage applies stricter criteria.

For each candidate case, collect the following minimum dataset:

• Animal identifier, species, breed, age, sex, and neuter status
• Weight and body condition score
• Drug name, formulation, dose, route, frequency, and duration
• Indication for treatment
• Concurrent medications and vaccinations within a defined window
• Relevant medical history, including prior adverse reactions
• Onset date and time relative to drug administration
• Clinical signs, physical examination findings, and clinicopathologic results
• Outcome, including recovery, euthanasia, or death
• Whether the event was reported spontaneously or detected through active surveillance

The completeness of this dataset determines the quality of the causality assessment. Missing data are common in spontaneous reports, and the protocol should state how missing fields will be handled, for example through exclusion from specific analyzes or through sensitivity analyzes that test different assumptions.

Causality Assessment

Causality assessment assigns a probability that the drug caused the observed event. No single algorithm is universally accepted in veterinary medicine, and the assessment should integrate several lines of evidence.

Temporal plausibility is the first criterion. The event must occur after drug administration, and the interval must be biologically plausible for the suspected mechanism. Hepatotoxicity can appear after days to weeks of treatment, whereas anaphylaxis typically occurs within minutes to hours. The review of drugs withdrawn from the market for hepatotoxicity identified inappropriate liver function test monitoring and non-compliance with treatment as factors that delayed recognition of serious injury Babai et al., hepatotoxicity withdrawal review. This finding applies directly to veterinary monitoring protocols: the absence of baseline and serial biochemistry values makes post hoc causality assessment unreliable.

Dechallenge and rechallenge provide strong evidence when available. Improvement after drug withdrawal supports causation, and recurrence after re-exposure is confirmatory. Rechallenge is rarely ethical in veterinary patients and should only be considered when the drug is essential and the event was mild and reversible.

Biological plausibility draws on pharmacologic mechanism, species-specific metabolism, and published evidence. Some species metabolise drugs differently, and a reaction documented in one species may not predict the reaction in another. The reverse is also true: absence of a reaction in laboratory animals does not exclude a reaction in the target species, a limitation recognized in human drug safety where animal models have failed to predict serious hepatotoxicity Babai et al., hepatotoxicity withdrawal review.

The final assessment should classify each case using a standard scale, for example certain, probable, possible, unlikely, or unassessable. The protocol must state the criteria for each category and require that two assessors work independently with adjudication of disagreements.

Signal Evaluation and Prioritization

A signal is a set of data that suggests a new potentially causal association or a new aspect of a known association. Signals require evaluation before they can be acted upon. The evaluation should consider the strength of the statistical association, the consistency across data sources, the biological plausibility, and the clinical severity of the event.

Prioritization determines which signals warrant regulatory action, label changes, or clinical guidance. High-priority signals involve fatal or irreversible events, events in young or breeding animals, events affecting food safety, or events with no alternative explanation. Lower-priority signals involve mild, reversible events with weak statistical support.

The evaluation should also consider the expected background rate of the event in the treated population. Some adverse events occur spontaneously in the species, and the signal must be interpreted against that baseline. For example, gastric dilatation-volvulus occurs in large-breed dogs without drug exposure, and a signal linking a drug to this event requires careful comparison with the background incidence.

Reporting and Dissemination

The study report must describe the methods in sufficient detail for replication. The ARRIVE guidelines specify the minimum information required for transparent reporting of animal research, including animal characteriztics, experimental procedures, and statistical methods ARRIVE guidelines 2.0. The EQUATOR Network provides a library of reporting guidelines that includes extensions relevant to observational pharmacoepidemiology EQUATOR Network reporting guidelines. Veterinary investigators should consult these resources when structuring their manuscripts.

The report should include a flow diagram showing the number of reports identified, excluded, and included at each stage. The reasons for exclusion must be stated. The results section should present the crude counts, the calculated rates or disproportionality scores, and the confidence intervals. The discussion should address the limitations, particularly the potential for reporting bias, confounding by indication, and incomplete data.

Findings that meet the threshold for a safety signal should be communicated to the relevant regulatory authority and to the marketing authorisation holder. In many jurisdictions, veterinarians have a professional obligation to report suspected adverse reactions. The report should also be disseminated to the veterinary community through peer-reviewed publication, because spontaneous reporting systems depend on clinician awareness and participation.

Recognized Complications and Failure Modes

Pharmacovigilance studies fail in characteriztic ways. Under-reporting remains the most pervasive limitation of spontaneous systems, because the denominator of actual adverse events is unknown and the numerator is incomplete. Detection bias arises when a product is new, heavily promoted, or subject to media attention, inflating report volumes independently of true risk. Stimulated reporting follows Dear Doctor letters, label changes, or regulatory announcements, creating transient spikes that can be misread as emerging signals.

Notoriety bias operates in the opposite direction: once a drug is suspected of causing a particular reaction, clinicians report that reaction preferentially, while other adverse events for the same drug go unreported. Protopathic bias occurs when the indication for treatment is an early manifestation of the outcome under study, so the drug appears causative when the disease itself produced the clinical sign. For example, a non-steroidal anti-inflammatory drug prescribed for lameness may be temporally associated with a subsequent diagnosis of neoplasia that was already present.

Survivor cohort effects distort long-term safety assessments. Animals that tolerate a drug continue therapy, while those that react are withdrawn, so chronic-dosing cohorts are enriched for tolerant individuals. This can mask cumulative toxicity, a problem recognized in human hepatotoxicity surveillance where delayed, irreversible liver injury escaped detection during clinical development safety data and withdrawal of hepatotoxic drugs. The same mechanism applies to veterinary products with delayed organ toxicity.

Detection of these failure modes requires structured interrogation of the data. Stratify reporting rates by time on market, by indication, and by reporter type. Plot cumulative report counts against time and inspect for step changes that coincide with external events. Compare reporting patterns for the suspect drug against a comparator class with similar indications. When a signal rests on a small number of reports, examine the original case narratives instead of coded terms alone.

Common Errors in Study Conduct

Inexperienced investigators frequently conflate statistical signal with causal effect. A disproportionate reporting ratio identifies an association worthy of evaluation, not a proven adverse reaction. The Multi-item Gamma Poisson Shrinker method used in human pharmacovigilance produces adjusted observed-to-expected ratios that rank associations, but these rankings require clinical review before any regulatory action atypical antipsychotics and pituitary tumor pharmacovigilance analysis. The same discipline applies in veterinary data.

Case definition errors are equally common. Investigators may accept a diagnosis recorded in the medical record without verifying that it meets prespecified criteria, or they may include cases with incomplete follow-up that cannot support a temporal relationship. Duplicate reports from different sources, such as the same event reported by both the owner and the attending veterinarian, inflate counts. Corrective action includes a formal deduplication protocol, a priori case definitions, and independent adjudication of ambiguous cases by a second reviewer.

A third recurring error is over-interpretation of denominator data. Practice-based electronic health records provide prescription counts, but not all dispensed drugs are administered, and not all administered drugs are dispensed through the practice. Studies that use sales data as a proxy for exposure should state this assumption explicitly and consider sensitivity analyzes that vary the exposure window.

Limitations of the Current Evidence

The veterinary pharmacovigilance evidence base is thinner than its human counterpart. Spontaneous reporting systems for animals capture only a fraction of events, and the quality of reports varies widely with reporter training and species. Production animal surveillance is complicated by herd-level treatment, where individual exposure is often unknown and multiple animals receive the same medication simultaneously. Food animal studies must also consider residue concerns and withdrawal periods, which have no analogue in companion animal or human pharmacovigilance.

Expert opinion still diverges on several points. There is no consensus on the minimum number of reports required to declare a signal in veterinary data, nor on the appropriate threshold for disproportionality statistics adjusted for the small size of veterinary databases. Some authorities advocate borrowing human thresholds, while others argue that veterinary reporting volumes are so low that any replicated association warrants investigation. The role of pharmacogenetics in veterinary adverse reactions is similarly unsettled, with breed-specific metabolic differences recognized in some species but poorly characterized in most.

Reporting standards for veterinary pharmacoepidemiology are also less developed than for clinical trials. The ARRIVE guidelines cover animal research reporting, and the EQUATOR Network catalogs reporting standards for human studies, but no equivalent veterinary-specific standard exists for observational safety studies ARRIVE reporting guidelines, EQUATOR reporting guideline library. Investigators should therefore follow the closest applicable human standard, such as STROBE for observational studies, and state explicitly where veterinary adaptations were made.

Escalation and Referral Pathways

Certain findings warrant escalation beyond the study team. A signal involving a food animal product with potential residue risk requires immediate consultation with the relevant regulatory authority, because public health implications extend beyond the treated animals. Suspected pharmacovigilance fraud, such as fabricated case reports or systematic under-reporting by a sponsor, should be referred to the competent authority without delay.

Specialist consultation is indicated when the adverse event falls outside the study team's expertise. Toxicological pathology review may be needed to adjudicate histopathology findings, and clinical pharmacology input can clarify dose-response relationships or suspected drug interactions. Teratology or reproductive toxicology expertise is advisable when signals involve pregnancy outcomes, given the recognized difficulty of interpreting such data from spontaneous reports GLP-1 agonist and SGLT2 inhibitor pregnancy safety evidence.

Regulatory reporting obligations differ by jurisdiction and product type. Investigators should determine at protocol stage which events require expedited reporting to the relevant authority and which are captured in periodic safety update reports. The World Organization for Animal Health maintains international standards for veterinary surveillance that can guide study design even where national requirements differ WOAH terrestrial animal health standards. When in doubt, consult the regulatory authority directly instead of relying on secondary interpretation.

Frequently Asked Questions

What Is the Minimum Viable Pharmacovigilance System for a Small Practice or Low-Resource Setting?

A small practice can begin with a structured spreadsheet or practice management system module that captures the minimum data elements: suspect product, batch number, species, signalment, reaction description, onset interval, outcome, and concurrent medications. Designate one team member as the local safety officer. Establish a monthly review habit and a clear escalation pathway for serious reactions. Use the AVMA practice resources for template adverse event forms and reporting workflows. Submit all suspected adverse reactions to the relevant national authority even when causality is uncertain, because spontaneous reports remain the backbone of signal detection. Prioritize data quality over data volume, and document any changes to the recording system so that later analyzes can account for them.

How Should I Handle a Suspected Adverse Reaction When the Product Label Provides No Relevant Warning?

Treat the absence of a label warning as an information gap, not evidence of safety. Record the reaction with the same rigour you would apply to a labelled event, including photographs, laboratory results, and histopathology where available. Search published literature and the MSD Veterinary Manual for species-specific reports of similar reactions. Contact the marketing authorisation holder directly, as they maintain internal safety databases that may contain unpublished cases. Report the event to the national pharmacovigilance center and request a literature search if their system permits it. For novel or severe reactions, consider publishing a case report, but follow the ARRIVE reporting guidelines if the case involves experimental procedures, and consult the EQUATOR Network library for the appropriate case report checklist.

What Are the Practical Differences Between Conducting Pharmacovigilance in Food-Producing Animals and Companion Animals?

Food animal pharmacovigilance carries additional consequences because a reaction may affect the safety of the human food supply. Withdrawal period adjustments, residue testing, and milk or egg discard decisions become immediate practical concerns. The WOAH terrestrial animal health standards frame the international expectations for veterinary drug oversight and residue control. Companion animal work relies more heavily on owner observation and electronic health record mining, whereas food animal work often draws on herd-level production data and slaughterhouse surveillance. Causality assessment in food animals must consider batch-level exposures across many animals, which can strengthen signal detection but also introduces clustering effects that complicate statistical analysis. Always document the production class and intended food use, because this determines which regulatory pathways apply.

How Do I Distinguish a True Drug Reaction From an Underlying Disease Progression in a Case I Am Reporting?

Apply a structured causality algorithm instead of clinical intuition. The WHO-UMC system and the Naranjo scale both work in veterinary contexts, though neither was validated for animals. Document the temporal relationship, the response to dechallenge and rechallenge where ethically feasible, and the plausibility of alternative explanations. Consider whether the reaction is a known dose-dependent effect, an idiosyncratic response, or a consequence of drug interaction. The hepatotoxicity literature illustrates how carefully this distinction must be drawn, since liver injury can be misattributed when monitoring is inadequate safety data and withdrawal of hepatotoxic drugs. When rechallenge is not possible, state this explicitly in the report and classify the case as possible instead of probable. Do not let diagnostic uncertainty prevent reporting, because the reporting system depends on receiving cases of all certainty levels.

What Records Must I Keep to Support a Future Pharmacovigilance Investigation?

Maintain the original medical record, the dispensing record, the batch number and expiry date, and any laboratory or imaging results contemporaneous with the reaction. Record the owner's description verbatim where possible, and note the time of drug administration relative to reaction onset. Preserve the product container and any remaining medication in a secure location in case the regulatory authority requests samples. Keep a log of all communications with the manufacturer and the regulatory body. For herd-level events, record the number of animals treated, the number affected, and the attack rate. The EQUATOR Network reporting guidelines provide a useful framework for deciding which data elements matter for later publication, even when you are collecting records primarily for regulatory purposes. Retain records according to your jurisdiction's requirements, which typically exceed the duration of the animal's life.

How Should I Explain a Suspected Adverse Drug Reaction to an Owner Without Causing Panic or Undermining Trust?

Lead with the facts: what was given, what happened, and what you are doing next. Distinguish between a known, expected reaction and a new or unexpected one, and be honest about uncertainty. Explain that reporting the reaction helps protect other animals, which reframes the event as constructive instead of threatening. Avoid defensive language and do not blame the manufacturer or the previous veterinarian. Describe the monitoring plan for the next 24 to 72 hours in concrete terms, including which signs warrant an immediate call. If a rechallenge is being considered, explain the rationale and the risks in plain language. The AVMA practice resources include client communication guidance that can help structure these conversations. Document the discussion in the medical record, including the owner's questions and your responses.

Related Clinical & Scientific Guides

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

References and Further Reading

• Safety data and withdrawal of hepatotoxic drugs.. 2021.
• Pharmacovigilance: An Overview.. 2018.
• Dapoxetine: a new option in the medical management of premature ejaculation.. 2012.
• Toxicological risks of Chinese herbs.. 2010.
• Effects of GLP-1 agonists and SGLT2 inhibitors during pregnancy and lactation on offspring outcomes: a systematic review of the evidence.. 2023.
• Atypical antipsychotics and pituitary tumors: a pharmacovigilance study.. 2006.
• ARRIVE Guidelines 2.0 for Reporting Animal Research. PLOS Biology, 2020.
• EQUATOR Network Reporting Guidelines. EQUATOR Network.
• MSD Veterinary Manual, Professional Edition. MSD Veterinary Manual.

Related Articles

• Appraising Diagnostic Accuracy Studies in Veterinary Medicine
• Diagnostic Test Accuracy Studies in Veterinary Medicine: Design and Reporting
• Evaluating Prognostic Factor Studies in Veterinary Medicine
• Observational Studies in Veterinary Medicine: Strengths and Weaknesses
• Conducting Qualitative Research in Veterinary Settings

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.