Real-World Evidence: From Pyramid to Amphitheatre
Posted on: Thursday 25 June 2026
Introduction
There has been a welcome shift away from over reliance on spontaneous reporting of adverse drug reactions (ADRs) to the use of multiple data sources to inform pharmacovigilance decisions. For decades, the hierarchical pyramid of evidence has been the cornerstone of evidence‑based medicine, placing meta‑analysis and randomised controlled trials (RCTs) at the top, and real‑world observational studies near the bottom.
This hierarchy serves an important purpose when assessing the efficacy of interventions, where randomisation and controlled conditions offer clear advantages. However, when the objective shifts to postmarketing drug safety, the pyramid begins to appear fundamentally unsound.
Real‑world patient populations are more diverse, more complex, and far less homogeneous than those enrolled in RCTs. When we consider how medicines are actually used in routine clinical practice and how safety issues emerge, it becomes clear that RCTs should not dominate. Drug safety requires a broader, more inclusive and non‑hierarchical approach to the evaluation of the evidence.
Clinical trials do not reflect real‑world patients
RCTs remain essential for establishing efficacy of therapeutic interventions, but their limitations for safety evaluation are well recognised. Clinical trials typically involve relatively small, carefully selected populations, usually excluding patients with comorbidities (such as renal and hepatic impairment or cancer) and vulnerable groups, including older adults and pregnant women.
In addition, trial follow‑up is usually short, limiting the ability to detect late ADRs. As RCTs are primarily designed to assess efficacy, they are therefore nearly always underpowered to study rare adverse events.
Limitations of RCTs and the role of observational studies
Spontaneous reporting systems have been the cornerstone of pharmacovigilance. They provide early warnings and help identify safety signals that were not detected or incompletely characterised during clinical development. However, spontaneous reports have well‑known limitations,
- under‑reporting
- lack of denominator data
- incomplete or variable information
While spontaneous reports are invaluable for signal detection, they rarely provide definitive answers for many safety concerns.
This is where real-world observational studies become not only useful, but essential for pharmacovigilance. Observational study designs, such as cohort, case-control and cross‑sectional studies, as well as registries frequently use electronic health records and other routinely collected data sources. These studies allow the monitoring of medicine safety in real-world clinical settings, often including thousands of patient‑years of observation.
Real‑world observational studies:
- examine causal
- detect rare and long‑term adverse drug
- characterise patterns of drug utilisation and the characteristics of patients who experience adverse drug reactions.
- link adverse drug reactions to genetic, biological, and environmental risk factors.
Like other research methods, observational studies have limitations, including bias, confounding, misclassification and missing information. However, guidelines such as STROBE (Strengthening the Reporting of Observational Studies in Epidemiology), ENCePP (European Network of Centres for Pharmacoepidemiology and Pharmacovigilance) and Good Pharmacoepidemiology Practice address these issues to a considerable extent. Understanding the residual impact of such limitations, even after adjustment is an additional safeguard when interpreting their results.
In many situations, conducting an RCT is neither feasible nor appropriate. Examples include comparisons of psychotropic medicines with talking therapies, or surgical procedures with pharmacological treatments. Furthermore, post‑marketing RCTs often struggle to recruit sufficient numbers to detect rare harms.
These realities clearly support the notion that the hierarchical pyramid of evidence model is inappropriate for evaluating safety in real-world settings. What is often considered the “highest” form of evidence may be the least informative for the safety questions.
The Rise of Real‑World Evidence
The contribution of real-world evidence (RWE) has expanded extensively over the last two decades, extending beyond its traditional roles in pharmacovigilance and outcomes research to all stages of the medicine lifecycle. Today, we have access to a richer array of real‑world data sources, including:
- electronic health records
- administrative claims
- disease and drug
- biobanks
- patient‑reported outcomes
- data from social media and wearables
Regulatory authorities, such the EMA, FDA and MHRA are increasingly integrating evidence from diverse data sources into their decision-making frameworks. Having spent many years working in pharmacovigilance, I view this growing reliance on RWE as a positive and necessary development.
From Pyramid to Amphitheatre
There is now substantial evidence that the most appropriate way to evaluate real-world evidence for pharmacovigilance, and beyond, is not through a hierarchy, but using an inclusive and integrative framework.
The amphitheatre concept recognises that:
- no single study design is inherently “best.”
- different methods answer different questions.
- the sources of supportive evidence vary depending on the objectives and the questions (for example, efficacy or safety)
- evidence should be judged on fitness for purpose, not on a predetermined rank.
- high‑quality decisions rely on the totality of evidence, not the apex of a pyramid.
This perspective aligns with long‑standing critiques of rigid evidence hierarchies, which can be misleading when applied indiscriminately. It also reflects the realities of pharmacovigilance, for example, a few well‑documented case reports may be more informative than a large RCT, and registry analyses can reveal patterns that a RCT has insufficient power to detect.
The amphitheatre approach does not diminish the value of RCTs. It places them in their proper context as a powerful tool among many. Adopting this approach acknowledges that real-world evidence is not an afterthought, but central to understanding how medicines perform once they move beyond the controlled environments of RCTs to routine clinical practice.
As pharmacovigilance evolves, so too must our expectations of what constitutes the best evidence. Regulatory frameworks in Europe and elsewhere already reflect this shift, increasingly integrating diverse data sources to support informed, real‑world decision‑making.
Conclusion
Drug safety today is more complex, dynamic and interdisciplinary than ever before. Continuing to rely on a rigid hierarchy of evidence, despite the nature of the data that inform pharmacovigilance decisions, is both inappropriate and potentially misleading.
By contrast, the amphitheatre (a circular, non‑hierarchical model of evidence evaluation) offers a more flexible, inclusive and informative approach to assessing safety outcomes. By embracing the totality of evidence, this model supports a more holistic framework for clinical decision‑making and the protection of public health.
Although initially focused on pharmacovigilance of pharmacotherapeutic interventions, this approach can be generalised to the evaluation of a broad range of therapeutic and diagnostic interventions. By evaluating all relevant evidence (including clinical trials, observational studies, spontaneous reports, registries and patient‑generated data), we are better positioned to make well‑informed decisions. The amphitheatre of evidence invites us to see the full landscape, rather than fixating on an illusory peak.
Reference
Shakir S, Lane S. BMJ Open 2025;15:e103538. doi:10.1136/bmjopen-2025-103538