Methods of ADR Detection in Pharmacovigilance

1. Spontaneous (Voluntary) Reporting Systems

• Healthcare providers, consumers, and manufacturers report suspected ADRs to regulatory bodies.
• Examples: FDA Adverse Event Reporting System (FAERS) in the US; Yellow Card Scheme (UK); WHO Programme for International Drug Monitoring (via VigiBase).
• Reporting by healthcare providers and consumers is voluntary; manufacturers are required to report.
• The FDA issues quarterly reports on new safety signals from FAERS, which currently receives >2 million reports/year.
• Case reports identified landmark ADRs such as thalidomide teratogenicity and cardiac valvular abnormalities from fenfluramine-phentermine.

• Massive underreporting - median underreporting rate across 12 countries was 94% (Hazell & Shakir, 2006).
• Denominator unknown - total number of patients exposed to a drug is not captured.
• Reports frequently incomplete - missing cause of death, past medical history, concomitant medications.
• Cannot detect small-to-moderate increases in risk for common events (e.g., MI risk increase of relative risk 1.2).
• Reporting patterns influenced by media coverage, litigation, and time since market entry.

2. Quantitative Signal Detection (Disproportionality Analysis)

• Contrasts the observed number of adverse event reports for a particular drug with the expected number based on the entire database.
• Methods include: Proportional Reporting Ratio (PRR), Reporting Odds Ratio (ROR), Bayesian Confidence Propagation Neural Network (BCPNN).
• Does not require knowledge of the true volume of drug use (the denominator).
• Signals identified require further evaluation through controlled studies before conclusions can be drawn.

3. Observational Epidemiological Studies

a) Cohort Studies

• Identify a population exposed to a drug and a comparison (unexposed) group, then follow both forward in time to measure outcomes.
• Can estimate incidence rates and relative risks.
• Example: statin-induced myopathy studied in a longitudinal cohort.
• Limitation: expensive, requires large sample sizes for rare ADRs.

b) Case-Control Studies

• Start with patients who have experienced a particular adverse event (cases) and compare their drug exposure history to matched controls who did not.
• Efficient for studying rare outcomes.
• Calculates odds ratios.
• Limitation: susceptible to recall bias and selection bias.

c) Case-Population Studies

• A hybrid approach comparing drug exposure in cases to a background population rate of exposure.

4. Prescription Event Monitoring (PEM)

• Used extensively in the UK (Drug Safety Research Unit).
• Prescriptions for a newly marketed drug are identified through dispensing records.
• A questionnaire ("green form") is sent to prescribing physicians asking for all events (desired or undesired) occurring after the prescription.
• Does not rely on a suspicion of causation - all events are collected.
• Provides both numerator and denominator data, allowing incidence rate calculation.
• Has been used to estimate rates of hepatic drug reactions and other organ-specific toxicity.

5. Intensive Hospital-Based Monitoring

• Systematic surveillance of all patients in a hospital ward or unit.
• All drugs administered and all adverse events experienced are recorded prospectively.
• Examples: Boston Collaborative Drug Surveillance Program.
• Advantage: high-quality, complete data with accurate denominators.
• Limitation: limited generalizability; hospitalized patients may differ from the general population.

6. Record Linkage Studies

• Links large administrative databases (e.g., pharmacy dispensing records, hospital admission records, death registries) to identify drug-event associations.
• Allows population-level pharmacoepidemiology without new data collection.
• Used in countries with national health records (UK, Canada, Scandinavia).
• Useful for identifying rare but serious events like drug-induced hepatotoxicity.

7. Randomized Controlled Trials (RCTs) with Safety Endpoints

• Post-approval Phase IV trials and large simple trials may be specifically designed to assess safety.
• Regulatory authorities (e.g., FDA) can require post-marketing safety studies as a condition of approval.
• Example: cardiovascular outcome trials mandated for newer antidiabetic agents.
• Advantage: eliminates confounding by design.
• Limitation: may be too short or too small to detect rare, delayed ADRs; selected patient populations may not reflect real-world use.

8. Meta-Analysis and Systematic Reviews

• Pooling data across multiple studies to detect ADRs that individual studies were underpowered to identify.
• Particularly useful when an ADR has a small-to-moderate effect size.
• Example: meta-analyses revealed the cardiovascular risks of rofecoxib (COX-2 inhibitors).

9. Pharmacogenomic / Genetic Screening Approaches

• Identifying genetic markers that predict susceptibility to specific ADRs.
• Example: HLA-B*5801 allele as a marker for severe cutaneous reactions (Stevens-Johnson syndrome) with allopurinol in certain populations.
• Allows prospective screening to prevent ADRs in high-risk individuals.
• Part of the expanding pharmacovigilance science alongside traditional epidemiological methods.

10. Big Data and Real-World Evidence (RWE)

• Electronic health records (EHRs), claims databases, patient registries, and wearables are increasingly used.
• Advanced analytics and machine learning applied to large datasets can detect safety signals earlier.
• The FDA's Sentinel System is a national active surveillance system using claims and EHR data across tens of millions of patients.
• Complements spontaneous reporting, especially for common events where traditional methods fall short.

Summary Table