Estimating Patient Exposure in Pharmacovigilance

A count of adverse-event reports has little meaning without context. One hundred reports may represent a very different safety picture in a medicine used by ten thousand patients than in one used by ten million. Pharmacovigilance therefore needs an estimate of exposure: how many people received the medicinal product, for how long, at what dose and, where relevant, in which populations.

During controlled clinical development, exposure can often be measured directly from treatment records. After authorisation, exact worldwide patient counts are rarely available. Medicines move through wholesalers, hospitals, pharmacies and different healthcare systems; prescribed doses vary; adherence is imperfect; packs can remain unused; and many commercial datasets measure supply rather than ingestion. Post-authorisation exposure is therefore usually an estimate constructed from imperfect but useful sources.

The objective is not false precision. A good exposure estimate is one whose method is appropriate to the question, whose assumptions are explicit, whose uncertainty is understood and whose limitations are carried into the interpretation of the safety data.

Purpose and Regulatory Context

Exposure estimates are used in periodic benefit-risk evaluation, development safety reporting, signal assessment, pharmacoepidemiology, risk-management planning and broader benefit-risk evaluation. Different contexts require different levels of precision.

PBRER requirements

ICH E2C(R2), the Periodic Benefit-Risk Evaluation Report guideline, expects both cumulative and interval patient exposure from marketing experience where possible. It states that the estimated number of patients exposed should be provided when possible, together with the method used. If patient numbers are unavailable, alternative measures such as patient-days or numbers of prescriptions may be used; only when these are unavailable may drug-sales measures such as tonnage or dosage units be presented. The guideline also notes that Defined Daily Dose methodology may be used to estimate exposure.

PBRER exposure should be stratified where relevant and feasible—for example by indication, sex, age, dose, formulation or region—particularly when those dimensions are important to the safety question.

DSUR requirements

ICH E2F requires the DSUR to describe cumulative subject exposure in the development programme, including numbers assigned to investigational product, placebo or comparators and available demographic subgroup information. It also expects explanation of the method used to estimate exposure and its limitations. If the investigational product is already marketed, the DSUR should include an estimate of cumulative marketing exposure based on the most recent PSUR/PBRER or another suitable source.

The distinction is important: clinical-trial exposure is usually counted from subject-level records; post-marketing exposure is often estimated indirectly.

What Exposure Actually Measures

Exposure can be expressed in several ways, each answering a different question.

Number of patients exposed

This is intuitive and useful when treatment is short or administered once or in discrete courses. It answers: approximately how many individuals received the product?

For chronic therapy, however, two patients treated for one month each and two patients treated for five years each still count as two patients. Patient count alone therefore ignores duration.

Patient-time

Patient-time combines the number of patients and duration of exposure. Common units include patient-days, patient-months and patient-years.

One patient treated for one year contributes one patient-year. Two patients treated for six months each also contribute one patient-year. This measure is useful when risk may depend on cumulative or prolonged exposure.

Patient-time derived from individual records is measured more directly than patient-time inferred from commercial sales.

Number of prescriptions or dispensings

Prescription or dispensing counts are often closer to actual utilisation than wholesale sales, but they still do not prove that a patient took the medicine. They can nevertheless provide useful exposure context and may be the best available metric in some markets.

Sales or units distributed

Sales data measure product movement through the supply chain, not patient treatment. They may be the only globally consistent source for a marketed medicine, but converting them to patients or patient-time requires assumptions about dose, pack use, treatment duration and wastage.

Defined Daily Dose

The WHO Defined Daily Dose (DDD) is the assumed average maintenance dose per day for a drug used for its main indication in adults. WHO explicitly describes the DDD as a technical unit of measurement, not necessarily the recommended or prescribed daily dose.

This makes DDD useful for standardised drug-utilisation comparisons, but it can be a poor proxy for true patient exposure when dosing is highly individualised, weight-based, indication-specific or intermittent. WHO does not assign DDDs to some major drug groups, including vaccines and many antineoplastic agents.

Choosing the Exposure Metric

The exposure metric should follow the safety question.

An acute reaction occurring immediately after a one-time treatment may be most interpretable against numbers of treated patients or doses administered. A cumulative toxicity emerging after years of treatment may require patient-years or duration-stratified exposure. A vaccine may be contextualised by doses or vaccination courses. An oncology medicine with weight- or surface-area-based dosing may require a product-specific method rather than DDD.

There is therefore no universally superior denominator. The key is to match the denominator to the biological and clinical mechanism of risk.

Sources of Exposure Data

No single source is best for every purpose. Exposure estimation often combines several sources whose strengths differ.

Clinical-trial records

Clinical trials provide the most direct exposure information because treatment assignment, dosing, start and stop dates, interruptions and follow-up are recorded prospectively. This allows precise counts and patient-time calculations within the study population.

Their limitation is representativeness. Trial populations may exclude older people, comorbid patients or concomitant therapies common in practice. Trial exposure is therefore highly accurate for the trial but not automatically representative of post-authorisation use.

Sales and distribution data

Sales data are broad and often globally available, making them practical for worldwide aggregate reporting. They can be converted to approximate treatment-days or patient counts using dose and duration assumptions.

However, the conversion can be distorted by:

• stock remaining in the supply chain;
• wastage or expiry;
• stockpiling;
• hospital inventory;
• multiple pack sizes and strengths;
• regional differences in prescribed dose; and
• treatment interruption or non-adherence.

The resulting number should therefore be labelled as an estimate rather than an observed count.

Prescription and dispensing data

Prescription and pharmacy-dispensing datasets can provide patient-level or transaction-level information, often with age, sex, indication or duration proxies. They may therefore support better stratification than sales data.

A prescription does not guarantee dispensing, and dispensing does not guarantee ingestion. Coverage may also be incomplete if patients obtain medicines outside the captured system.

Claims and electronic health records

Claims and electronic health record data can provide longitudinal exposure, treatment switching, comorbidity and outcome context. These sources are particularly valuable for pharmacoepidemiological studies where denominators and comparator cohorts are required.

Their validity depends on data provenance, coding, completeness, linkage, care-setting coverage and the operational definition of exposure. A large database does not eliminate these limitations.

Registries and PASS

Registries and post-authorisation safety studies can provide detailed exposure information for selected populations, including pregnancy, rare diseases or risk-minimisation programmes. Their strength is specificity; their limitation is usually narrower population coverage.

Converting Product Use Into Exposure

A sales-derived estimate often starts with the amount of product distributed and converts that quantity into treatment time or patient numbers.

A simplified approach is:

treatment-days = total quantity distributed ÷ assumed average daily dose

and, where an average treatment duration is defensible:

estimated patients = treatment-days ÷ assumed average treatment duration.

For chronic treatment, treatment-days can instead be converted to patient-years:

patient-years = treatment-days ÷ 365.25.

These equations are simple; choosing valid assumptions is not.

Average daily dose

The average daily dose may come from prescribing information, market research, prescription databases or observed utilisation studies. If several indications use materially different doses, one pooled daily-dose assumption may be misleading.

Treatment duration

Duration may be known for fixed-course therapies but highly uncertain for chronic treatment. Discontinuation, persistence, dose interruptions and switching can all affect the estimate.

Pack and strength mix

When several strengths are marketed, the conversion should account for the actual mix sold. Treating every tablet as equivalent can distort exposure if strengths differ substantially.

Regional variation

A global conversion factor can be inappropriate when prescribing patterns differ between countries. Where data permit, regional or indication-specific assumptions may improve validity.

Exposure in Special Populations

Aggregate exposure is often insufficient when a safety concern is concentrated in a subgroup. ICH E2C(R2) specifically encourages presenting available exposure information for relevant special populations and stratified categories.

Important subgroups may include:

• paediatric or elderly patients;
• pregnant or breastfeeding patients;
• renal or hepatic impairment;
• particular indications;
• specific formulations or routes;
• high-dose or long-duration users; and
• regions with materially different treatment practice.

The purpose is not to create subgroup estimates for every conceivable characteristic. Stratification should be driven by the product and safety question.

Reporting Rates Versus Incidence

One of the most important interpretive distinctions is between a spontaneous reporting rate and a true incidence rate.

A spontaneous reporting rate may be calculated as:

number of reported cases ÷ estimated exposure.

This can be useful for internal trend description, but the numerator is incomplete because not every event is reported, and the denominator is often estimated. Reporting probability can also change over time because of publicity, regulatory action, launch effects or stimulated reporting.

A spontaneous reporting rate therefore should not be described as incidence unless the underlying data source and study design actually capture events and exposure sufficiently to estimate incidence.

Comparing reporting rates over time

Changes can still be informative if methods are stable. For example, an increase in reports accompanied by a much larger increase in exposure may imply a different pattern from an increase in reports with stable exposure.

However, a methodology change in the denominator can create an artificial trend. Aggregate reports should therefore explain material changes in exposure-estimation method between reporting periods.

Handling Uncertainty

Exposure estimates should make uncertainty visible rather than hiding it behind a precise-looking number.

Useful approaches include:

• stating the source and data cut-off;
• documenting dose and duration assumptions;
• explaining whether the estimate measures distribution, dispensing or actual treatment;
• providing ranges or sensitivity analyses where assumptions materially affect the result;
• describing missing countries or incomplete data; and
• separating measured from modelled exposure where possible.

For example, if average treatment duration plausibly ranges from three to six months, showing how the estimated number of patients changes across that range may be more scientifically informative than selecting one value without explanation.

Practical Application in Aggregate Reports

Exposure should be presented in a way that allows the reader to understand both the denominator and the uncertainty around it. A useful aggregate-report presentation normally states the period covered, interval and cumulative estimates where applicable, the source of the data, the calculation method, important assumptions and any material limitations.

PBRER presentation

For marketing experience, ICH E2C(R2) expects interval and, where possible, cumulative exposure. The estimate should be stratified where relevant to the product and safety questions. If the reporting period includes a major change in indication, formulation, geography or utilisation, the exposure section should explain how that change affects interpretation.

A strong PBRER does not merely state "approximately 2.4 million patients were exposed". It explains enough of the method for a reviewer to understand what that number means—for example whether it was derived from units sold, prescriptions, patient-level dispensing data or a combination of sources.

DSUR presentation

For the development programme, cumulative subject exposure should remain aligned with the clinical-trial inventory and treatment allocation. Where treatment remains blinded, estimates may be based on the randomisation scheme as described in ICH E2F. The method and its limitations should be stated.

Where the investigational product is already marketed, the DSUR can use the most recent post-marketing exposure estimate or another suitable source, but the relationship between development and marketed exposure should remain clear.

Quality and Governance

Exposure estimation frequently draws data from commercial, epidemiological, finance, supply-chain, clinical or vendor sources outside the core PV database. Governance should therefore establish ownership for the method and for the source data on which it depends.

The organisation should be able to explain:

• where the source data came from;
• which period and territories are included;
• who owns and verifies the calculation;
• how assumptions were selected;
• how changes in methodology are controlled;
• how corrections to source data are handled; and
• where the calculation and rationale are retained.

The exact control model is organisational practice rather than a universal prescribed regulatory template.

Potential Failure Modes

The following are illustrative failure modes rather than reported inspection findings.

Sales are treated as patients

A unit sold is not a patient treated. Without dose and duration assumptions, sales volume should not be converted directly into patient numbers.

A single global dose assumption is used despite major regional differences

This can materially distort exposure when indications, body size, prescribing practice or pack strengths differ across markets.

DDD is treated as the prescribed dose

WHO defines DDD as a technical unit for drug-utilisation measurement. It may differ substantially from actual prescribed doses and is not assigned to all drug classes.

Reporting rate is labelled incidence

Spontaneous reports are incomplete and subject to reporting bias. Dividing them by estimated exposure does not create a population-based incidence rate.

The denominator method changes silently

An apparent safety trend may reflect a new commercial data source or revised dose assumption rather than a real change in product use. Material methodology changes should be visible to the reviewer.

Subgroup exposure is unavailable but subgroup risk is interpreted quantitatively

If a safety concern is concentrated in elderly patients, for example, a worldwide total-exposure denominator may not support a subgroup rate comparison. The limitation should be stated rather than hidden.

Inspection and Assessment Considerations

A regulator, inspector or internal reviewer may test whether exposure estimates are scientifically supportable and reproducible rather than whether a specific software tool or formula was used.

Illustrative questions include:

• What data source underlies this exposure estimate?
• Does it represent product distributed, dispensed or actually administered?
• Which dose and treatment-duration assumptions were used?
• Are those assumptions appropriate across indications and regions?
• How are pack strengths and formulation differences handled?
• How was patient-time calculated?
• Why was DDD used, and is it appropriate for this product?
• Can the interval and cumulative figures be reconciled with prior reports?
• What changed when the methodology changed?
• Are subgroup exposure estimates available for the safety issue being assessed?
• Has a spontaneous reporting rate been clearly distinguished from incidence?

These are potential review questions, not a fixed inspection checklist.

Practical Checklist

Before finalising a material exposure estimate, the team should be able to confirm that:

• the scientific purpose of the denominator is clear;
• the best reasonably available data source has been selected;
• the estimate is labelled as measured or modelled appropriately;
• dose, duration and pack assumptions are documented;
• regional or indication-specific variation has been considered;
• patient counts and patient-time are not used interchangeably without explanation;
• DDD is used only where appropriate and is not described as actual prescribed dose;
• interval and cumulative exposure are distinguished where required;
• special-population exposure is considered when relevant;
• spontaneous reporting rates are not presented as incidence;
• limitations and missing data are explicit;
• material methodology changes are documented; and
• the calculation can be reconstructed from retained source data and assumptions.

Key Takeaways

Patient exposure is the denominator that gives aggregate safety data context, but post-authorisation exposure is usually an estimate, not an exact count. The quality of that estimate depends less on mathematical complexity than on whether the source and assumptions are appropriate to the product and safety question.

ICH E2C(R2) expects estimated patient exposure in PBRERs where possible, together with the method used, and allows alternative measures such as patient-days, prescriptions, sales units and DDD-derived estimates when direct patient numbers are unavailable. ICH E2F similarly requires cumulative development exposure in the DSUR and explanation of the estimation method and its limitations.

Patient counts, prescriptions, sales, patient-years and DDDs answer different questions. A denominator should therefore be selected according to treatment pattern and the biology of the risk, not simply because one commercial dataset is convenient.

The most important interpretive safeguard is to distinguish reporting rates from incidence. Spontaneous-report numerators and estimated exposure denominators can support contextual interpretation and trend assessment, but they do not by themselves provide population incidence or causal risk estimates.

A robust exposure section is transparent enough for another reviewer to understand what was measured, what was assumed, how the estimate was derived and how uncertainty affects the safety conclusion.

References

1. International Council for Harmonisation. ICH E2C(R2): Periodic Benefit-Risk Evaluation Report (PBRER). See section 3.5 on estimated exposure and section 3.5.2 on marketing experience. https://database.ich.org/sites/default/files/E2C_R2_Guideline.pdf
2. International Council for Harmonisation. ICH E2F: Development Safety Update Report (DSUR). See section 3.6 on estimated cumulative exposure. https://database.ich.org/sites/default/files/E2F_Guideline.pdf
3. World Health Organization. Defined Daily Dose (DDD): definition and general considerations. https://www.who.int/tools/atc-ddd-toolkit/about-ddd
4. World Health Organization. ATC/DDD methodology. https://www.who.int/toolkits/atc-ddd-toolkit/methodology
5. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module VII – Periodic safety update report. Available from the EMA GVP collection: https://www.ema.europa.eu/en/human-regulatory-overview/post-authorisation/pharmacovigilance-post-authorisation/good-pharmacovigilance-practices-gvp
6. European Medicines Agency. Guideline on good pharmacovigilance practices (GVP) Module IX – Signal management (Rev. 1). Available from the EMA GVP collection.

Regulatory Note

ICH E2C(R2) and ICH E2F provide internationally harmonised regulatory guidance for PBRER and DSUR exposure reporting. The choice of commercial data source, dose-conversion method, DDD use, patient-time model, sensitivity analysis and governance workflow is product- and context-dependent unless an applicable authority or procedure specifies otherwise. WHO's DDD is a technical unit of drug-utilisation measurement and should not be presented as the recommended or actual prescribed daily dose. Spontaneous reporting rates calculated with estimated exposure should not be described as incidence rates unless the underlying data and study design support incidence estimation.