Drug Utilisation Studies in Pharmacovigilance

Purpose and Scope

A drug utilisation study (DUS) examines how a medicinal product is prescribed and used in routine clinical practice. The central object of study is exposure: which patients receive the medicine, for what indication, at what dose, for how long, in what sequence with other treatments, and under what monitoring or prescribing conditions.

That makes DUS different from an aetiological safety study. A cohort or case-control study may ask whether exposure causes or increases the risk of an outcome. A DUS may instead ask whether the medicine is reaching patients for whom a particular risk is relevant, whether contraindicated or off-label use is occurring, or whether a risk-minimisation intervention has changed prescribing behaviour.

The distinction is not absolute. A DUS can supply denominator information, define the exposed population, identify confounders for a later safety study and, when its main aim is to add knowledge about safety or the effectiveness of risk-minimisation measures, may meet the definition of a post-authorisation safety study (PASS). The regulatory classification therefore follows the study objective, not the label attached to the protocol.

Why Utilisation Matters to Pharmacovigilance

Safety data cannot be interpreted well without understanding exposure. Ten reports in ten thousand exposed patients have a different epidemiological context from ten reports in ten million, although spontaneous-reporting data alone still cannot usually be converted into incidence because reporting is incomplete and selective.

Utilisation information also changes the meaning of a safety concern. If a medicine with renal toxicity is increasingly used in patients with renal impairment, the risk-management problem is not merely whether the adverse reaction exists. It also includes whether the exposed population is shifting toward patients with greater susceptibility and whether prescribing controls are functioning as intended.

DUS can therefore support pharmacovigilance by characterising:

• treated populations and clinically important subgroups;
• indications and patterns of off-label use;
• dose, duration, switching and treatment persistence;
• co-medication and combinations relevant to interaction risk;
• adherence to contraindications, testing or monitoring recommendations;
• prescribing changes after regulatory action;
• implementation of additional risk-minimisation measures;
• the size and distribution of exposed populations for safety-study planning.

Regulatory Framework

GVP Module VIII

EU GVP Module VIII on post-authorisation safety studies includes drug utilisation studies among recognised pharmacoepidemiological designs. Its appendix describes DUS as studies of prescribing and use in routine practice, including populations that may be under-represented in randomised trials. It also explains that a DUS whose main aim is to add knowledge about safety or the effectiveness of risk-minimisation measures may be classified as a PASS.

This is an important qualification. Not every DUS is automatically a PASS. A market-description exercise or a study undertaken for a non-safety purpose does not become a PASS merely because it uses prescription data. Conversely, a study designed to determine whether an imposed risk-minimisation measure is changing prescribing behaviour may fall squarely within the PASS framework.

GVP Module XVI and effectiveness evaluation

GVP Module XVI Rev. 3, effective from August 2024 for the circumstances described in that guidance, places greater emphasis on a planned implementation pathway for risk-minimisation measures and on prospective effectiveness evaluation. Module XVI Addendum II describes methods for evaluating implementation, knowledge, behaviour and health outcomes. Drug utilisation studies are particularly relevant to the behavioural-change layer because prescribing, dispensing, testing and monitoring actions can often be measured from healthcare data.

For embryo-fetal risks, the 2025 Module XVI Addendum I specifically recognises DUS as a possible method for examining prescribing in females of reproductive age and adherence to risk-minimisation actions over time or before and after an intervention.

Legally binding requirements versus guidance

The legal obligations governing an individual study depend on why and how it is conducted. A PASS may be imposed as a condition or may be conducted voluntarily, and imposed non-interventional PASS are subject to specific legal and procedural requirements. GVP Modules VIII and XVI provide regulatory guidance on implementing those requirements and on good scientific practice.

An internal company decision that every DUS requires a particular committee, template or QPPV signature is therefore an operational control, not an EU-wide legal rule unless a specific legal or regulatory condition makes it so.

Formulating the Research Question

A useful DUS begins with a decision-relevant question. Broad aims such as “describe utilisation” often produce large tables without resolving the regulatory uncertainty.

A stronger question specifies the population, exposure pattern, action or behaviour of interest, time period and comparison where relevant. Examples include:

• What proportion of new users had the required laboratory test before treatment initiation?
• Did prescribing to a contraindicated subgroup decrease after implementation of an additional risk-minimisation measure?
• Are patients receiving longer treatment durations than those described in the authorised product information?
• Has concomitant use with an interacting medicine changed after a safety communication?

The question determines the data needed and the design that can answer it. A prescription database may show that a medicine was dispensed, but not why it was prescribed. Claims data may identify diagnoses imperfectly. Hospital records may contain rich clinical detail but capture only one part of the patient's care pathway. The study should therefore be designed around what the data can validly measure rather than around what variables happen to be available.

DUS, Consumption Studies and Aetiological Studies

These related approaches answer different questions.

A DUS may precede an aetiological study because it clarifies the exposed population, potential comparators and likely confounders. It may also follow regulatory action to determine whether clinical practice changed. These functions are complementary rather than interchangeable.

Study Designs and Data Sources

A drug utilisation study is defined more by its question than by one fixed design. Cross-sectional analyses can describe who is using a medicine at a particular time. Repeated cross-sectional analyses can show whether prescribing changes after regulatory action. Cohort designs can follow new users to characterise persistence, switching, treatment duration or monitoring. Interrupted time-series designs can be useful when the question is whether an intervention changed an established prescribing trend rather than merely whether use was different before and after a single date.

The data source must be capable of measuring the behaviour of interest. Prescription and dispensing databases are strong for treatment initiation and refill patterns but may not prove that a medicine was taken. Claims data can cover large populations but may contain limited clinical detail. Electronic health records may provide diagnoses, laboratory results and clinical context, while registries can offer disease-specific depth. Purpose-built surveys or chart review may be needed where routine data do not capture knowledge, counselling or clinical reasoning.

Exposure definitions

The study should define whether it is examining prescriptions written, medicines dispensed, medicines administered or medicines apparently taken. These are not interchangeable. A prescription indicates an intention to treat; dispensing confirms supply; administration data can show actual delivery in some settings; adherence measures infer use from patterns rather than directly observing ingestion.

New-user definitions are particularly important when studying initiation requirements. If a test is required before the first dose, a prevalent-user cohort can obscure whether the test occurred before treatment began. Washout periods used to identify new users should therefore be clinically and data-source appropriate.

Population and denominator

A useful DUS defines the eligible population before calculating proportions. The denominator for “percentage appropriately monitored” should not silently exclude patients with missing data unless that exclusion is justified. Likewise, the denominator for off-label use should reflect the population in whom indication can actually be classified.

Population stratification may be clinically essential. Age, sex, pregnancy potential, renal or hepatic impairment, concomitant medicines, prior disease and prescriber specialty can all alter the interpretation of utilisation. GVP Module VIII specifically recognises DUS as a way of characterising populations that may have been under-represented in randomised trials.

Measuring Utilisation

Initiation, prevalence and incidence of use

Prevalence of use describes how many people are using a medicine during a defined period. Incidence of use usually refers to treatment initiation among people at risk of starting therapy. These measures answer different questions. A stable prevalence can coexist with falling initiation if existing users remain on treatment for long periods.

Dose, duration and treatment patterns

Dose should be assessed using a clinically meaningful measure. Defined Daily Doses can be useful for population-level consumption comparisons, but the WHO DDD is a technical unit and is not necessarily the prescribed or recommended dose for an individual patient. Patient-level studies may instead require prescribed daily dose, administered dose, cumulative dose or dose intensity.

Persistence describes continuation of therapy over time. Adherence describes how closely medicine use follows the intended regimen. Switching, augmentation and discontinuation may be important when a safety action changes the preferred place of a medicine in therapy.

Indication and off-label use

Off-label use cannot be inferred reliably from absence of a diagnosis code alone. The study should consider coding completeness, timing of diagnoses, authorised indications during the study period and whether the data can distinguish treatment from diagnostic work-up or prophylaxis. Where indication is central to the question, validation or sensitivity analyses may be necessary.

DUS for Risk-Minimisation Effectiveness

Risk-minimisation effectiveness evaluation should begin by specifying the intended change. If the intervention aims to prevent initiation in a contraindicated group, the relevant endpoint is prescribing behaviour. If it aims to ensure laboratory monitoring, the endpoint is completion of that monitoring within the required clinical window. If it aims to reduce fetal exposure, measures may include prescribing to people with reproductive potential, pregnancy testing, contraceptive-related processes where measurable and pregnancy exposure outcomes.

GVP Module XVI Rev. 3 and Addendum II separate the implementation pathway from the effectiveness question. Distribution of a tool is not evidence that the intended behaviour changed. A high distribution rate may coexist with poor clinical uptake. Conversely, a behavioural improvement cannot automatically be attributed to the intervention if other regulatory actions, guideline changes, media attention or secular prescribing trends occurred at the same time.

Before-and-after studies

A simple pre/post comparison is easy to understand but vulnerable to confounding by time. If prescribing was already falling before the intervention, the post-intervention decline may not represent an intervention effect. Interrupted time-series analysis can estimate whether there was a change in level or trend while accounting for the pre-existing trajectory, provided enough observations are available before and after the intervention.

Comparison groups

Where feasible, an external or internal comparator can strengthen interpretation. A comparator medicine unaffected by the regulatory action may help distinguish a product-specific change from a broader shift in disease management. The comparator must nevertheless be clinically and temporally appropriate; an unsuitable comparator can create more bias than no comparator.

Bias, Missingness and Misclassification

DUS are observational studies and inherit the limitations of their data. Common sources of bias include incomplete capture of prescriptions outside the database, missing laboratory results performed in another system, misclassification of indication, incomplete recording of pregnancy status, differential follow-up and changing coding practices.

A robust protocol anticipates these limitations. It defines operational algorithms before analysing outcomes, validates important variables where feasible, uses sensitivity analyses to test alternative assumptions and distinguishes “not recorded” from “did not occur” unless the data source supports that inference.

For pharmacovigilance purposes, the central question is not whether the database is large but whether the measurement is valid enough to support the decision being made.

Interpretation in Benefit-Risk Assessment

Utilisation data can change the practical importance of a safety concern even without changing the biological risk itself. A rare adverse reaction may have greater public-health impact if exposure expands rapidly. Conversely, a regulatory restriction may reduce population exposure and therefore expected harm while leaving individual susceptibility unchanged.

DUS findings should therefore be interpreted alongside clinical safety evidence. They can quantify who is exposed and whether intended controls are functioning, but they usually cannot by themselves establish causality for an adverse reaction. Their greatest value is often in connecting pharmacovigilance evidence with actual clinical use.

Practical Implementation and Governance

A DUS should be governed in proportion to its regulatory purpose. The protocol should state the question, population, data source, exposure and outcome definitions, analysis plan, limitations and intended interpretation. Where the study is a PASS, the applicable legal and procedural requirements for PASS must be followed; where it is not, those requirements should not be imported merely because the methodology resembles a PASS.

Operationally, responsibilities should be clear for protocol approval, data access, statistical programming, medical and epidemiological review, quality control, interpretation, reporting and regulatory submission where applicable. Version control should preserve the relationship between the approved protocol, analysis specifications, datasets, outputs and final report.

Traceability of decisions

An experienced reviewer should be able to reconstruct why each major analytical choice was made. This includes cohort entry rules, exclusion criteria, coding algorithms, exposure gaps, risk windows, handling of missing data and sensitivity analyses. If an endpoint definition changes after seeing the data, the rationale should be documented and its impact on interpretation considered explicitly.

Interfaces with the pharmacovigilance system

DUS findings may affect the RMP, signal evaluation, PSUR/PBRER, safety communication, product information or additional risk-minimisation strategy. The study process should therefore include a route for escalating findings that may alter the benefit-risk profile or indicate that an existing risk-minimisation measure is not working as intended.

The QPPV's role is one of oversight within the pharmacovigilance system rather than necessarily authorship of every protocol or analysis. The organisation should be able to demonstrate how relevant DUS findings reach the people responsible for benefit-risk evaluation and regulatory action.

Special Situations

Multi-country studies

Cross-country DUS can improve generalisability but create comparability problems. Coding systems, prescribing pathways, reimbursement rules, access to laboratory data and clinical guidelines may differ. Harmonised definitions should therefore be tested against local data realities rather than imposed mechanically.

Product launches and rapidly changing use

Early post-launch utilisation can change quickly. Short observation windows may over-represent early adopters or specialist settings. Repeated analyses can be more informative than a single snapshot where uptake is evolving.

Pregnancy and reproductive-risk programmes

For embryo-fetal risk minimisation, GVP Module XVI Addendum I recognises DUS as one possible way to assess prescribing and adherence to relevant risk-minimisation actions. The feasibility of measuring pregnancy testing, contraception-related processes or pregnancy exposure depends heavily on the data source. Absence of a coded test or counselling record should not automatically be treated as non-compliance if care can occur outside the captured system.

Potential Failure Modes

The most important DUS failures are methodological rather than cosmetic. Illustrative failure modes include:

• treating dispensing as proof of ingestion;
• interpreting “not recorded” as “not done” without validating data capture;
• using an endpoint that does not correspond to the risk-minimisation objective;
• declaring an intervention effective from distribution metrics alone;
• ignoring secular trends in a simple pre/post comparison;
• using an unsuitable comparator;
• changing definitions after reviewing results without transparent documentation;
• describing a DUS as a PASS or non-PASS without considering its primary objective and regulatory status;
• presenting association or utilisation patterns as proof of clinical causality.

These are not invented inspection findings. They are methodological failure modes that can undermine the reliability and regulatory usefulness of the study.

Inspection and Review Considerations

An inspector or regulatory reviewer could reasonably examine whether the study was capable of answering its stated question, whether key decisions were prospectively defined, whether deviations were controlled, whether analysis was reproducible and whether important findings entered the pharmacovigilance decision process.

Illustrative questions include:

• Why was this data source considered adequate for the endpoint?
• How was new use distinguished from prevalent use?
• Can the organisation show that laboratory or indication data are sufficiently complete for the conclusion drawn?
• Were important protocol deviations or post hoc analyses documented?
• If the DUS evaluated risk minimisation, what was the predefined success criterion and how was it justified?
• How were the findings considered in the RMP or wider benefit-risk assessment?

The strength of the inspection evidence lies in traceability from the regulatory or scientific question to the data, analysis, interpretation and resulting action.

Key Takeaways

Drug utilisation studies describe how medicines are used in routine practice. They are especially valuable when pharmacovigilance needs to understand who is exposed, whether use is consistent with authorised conditions, how prescribing changes after regulatory action and whether risk-minimisation measures have changed behaviour.

A DUS is not automatically a PASS. Its regulatory classification depends on its objective and context. Good DUS practice therefore begins with a decision-relevant question, uses data capable of measuring that question, anticipates bias and misclassification, and connects findings to the wider pharmacovigilance system.

References

1. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module VIII – Post-authorisation safety studies (Rev. 3). EMA/813938/2011 Rev. 3. Legal effective date 13 October 2017.
2. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module XVI – Risk minimisation measures (Rev. 3). EMA/204715/2012 Rev. 3, 26 July 2024. Legal effective date 6 August 2024.
3. European Medicines Agency. GVP Module XVI Addendum II – Methods for evaluating effectiveness of risk minimisation measures. EMA/419982/2019. Legal effective date 6 August 2024.
4. European Medicines Agency. GVP Module XVI Addendum I – Risk minimisation measures for medicinal products with embryo-fetal risks. EMA/608947/2021, 22 August 2025. Legal effective date 29 August 2025.
5. World Health Organization Collaborating Centre for Drug Statistics Methodology. ATC/DDD methodology. Current online guidance.

Regulatory Note

This article distinguishes EU legal obligations from GVP guidance and recommended operational practice. Whether a particular drug utilisation study is subject to PASS-specific legal and procedural requirements depends on its purpose, design and regulatory status. Current legislation, the applicable marketing-authorisation conditions and current EMA/HMA procedural guidance should be checked for each study.