Pragmatic vs Explanatory Trials

Explanatory trials test efficacy: selected patients, strict protocol, ideal delivery — maximising internal validity. Pragmatic trials test effectiveness: broad inclusion, a usual-care comparator and real-world delivery — maximising external validity. The PRECIS-2 wheel maps where a trial sits on that spectrum across nine design domains.

A single design axis. The explanatory pole optimises every choice — who’s enrolled, how tightly the protocol is followed, how the comparator behaves — to give the intervention its best chance (efficacy). The pragmatic pole deliberately mirrors messy routine practice (effectiveness). PRECIS-2 scores nine such domains to locate a trial on the line.

Ask what question the design answers. A result from the explanatory (left) end tells you the ceiling of benefit in ideal hands; a result from the pragmatic (right) end tells you what to expect in your actual ED. Judge each by the standards of its own end of the spectrum.

A glowing efficacy result need not survive contact with real-world practice — broader patients, imperfect adherence, and a usual-care arm that is itself improving can all shrink or erase the effect. Knowing where a trial sits tells you how far you can generalise it.

• Explanatory = efficacy, internal validity, “CAN it work?”
• Pragmatic = effectiveness, external validity, “DOES it work?”
• PRECIS-2 wheel scores the design across 9 domains

Early goal-directed therapy for sepsis — the single-centre, explanatory Rivers trial (NEJM 2001) suggested protocolised EGDT cut mortality. The large pragmatic multicentre trilogy — ProCESS (NEJM 2014), ARISE (NEJM 2014) and ProMISe (NEJM 2015) — then found EGDT no better than usual care in routine practice. An efficacy signal from one selected centre did not transfer: by the 2010s “usual care” had itself absorbed early recognition, fluids and antibiotics, so the protocol’s incremental benefit largely vanished — the classic efficacy-to-effectiveness gap.

• Conflating efficacy (can it work?) with effectiveness (does it work in practice?).
• Misreading a “usual care” comparator — it moves over time and is the right pragmatic benchmark, not a flaw.
• Over-generalising an explanatory result to populations and settings it never studied.