Explanatory Trial

Explanatory Trial Design, Efficacy and Economic Evaluation

An explanatory trial asks whether a treatment can work when everything is arranged in its favour, with chosen patients, expert clinicians and closely watched adherence. Its opposite, the pragmatic trial, asks whether the treatment works in the routine care where it would actually be used. The difference matters to health economists because cost-effectiveness analysis guides decisions about routine care, where an effect measured under ideal conditions can overstate what a health system gets for its money. Since almost all trials have historically measured efficacy, analysts often need to spot explanatory features and adjust for them in a decision model.

Schwartz and Lellouch and the two attitudes to trials

Schwartz and Lellouch introduced the distinction in 1967 in the Journal of Chronic Diseases, and the paper was republished in the Journal of Clinical Epidemiology in 2009. They argued that most therapeutic trials are poorly formulated from the start, because they may be aimed at either of two radically different kinds of problem, and that the resulting ambiguity runs through the choice of treatments, subjects, comparisons and assessment.

They used "explanatory" for trials that test a causal hypothesis, for example that a drug produces a particular biological effect, and "pragmatic" for trials that help choose between options for care. The CONSORT extension for pragmatic trials notes that they saw pragmatism as an attitude to design, not a property of the trial, and quotes their view that the pragmatic approach would often have been more justifiable.

The distinction maps onto Archie Cochrane's three questions as summarised by Brian Haynes. Efficacy asks whether an intervention does more good than harm under ideal circumstances ("Can it work?"), effectiveness asks the same under usual circumstances ("Does it work in practice?"), and efficiency relates the effect to the resources consumed ("Is it worth it?"). Haynes notes that efficacy and effectiveness trials had also been called explanatory and management trials. An explanatory trial answers the first question, and economic evaluation needs an answer to the third.

Design choices that make a trial explanatory

A trial becomes explanatory through a series of design decisions, each of which raises the chance of detecting whatever benefit the treatment has. The 2009 PRECIS paper by Thorpe and colleagues set out the extreme explanatory and pragmatic choices for ten design domains, and the table summarises them.

The CONSORT extension gives a real explanatory example. In the North American Symptomatic Carotid Endarterectomy Trial (NASCET), surgeons had to have performed at least 50 carotid endarterectomies in the previous 24 months with a postoperative stroke or death rate below 6%, and the main interest lay in patients with severe stenosis, thought most likely to benefit. Explanatory designs also rely on blinding to stop belief in the treatment confounding the causal link, whereas pragmatic trials may let such beliefs contribute to the measured effect, as they would in practice.

Placing a trial on the PRECIS-2 wheel

The two labels mark the ends of a continuum, and one trial can sit at different points in different domains. Thorpe and colleagues judged a purely explanatory trial to be probably impossible, since no patient complies perfectly for ever and even the most skilled surgeon occasionally slips. PRECIS shows these positions at the design stage as spokes of a wheel, with the explanatory pole at the hub and the pragmatic pole at the rim.

PRECIS-2, published by Loudon and colleagues in 2015 with input from more than 80 trialists, clinicians and policymakers, uses nine domains. These are eligibility, recruitment, setting, organisation, flexibility of delivery, flexibility of adherence, follow-up, primary outcome and primary analysis. Each is scored from 1 (very explanatory) to 5 (very pragmatic). The tool deals only with applicability, meaning how easily results can be used in the intended settings, and leaves internal validity to other tools.

Both approaches have their place, and the aim is a design matched to purpose. Thorpe and colleagues describe the usual sequence, in which an early trial informing a research decision about a new drug is more explanatory and a later trial meant to guide clinicians or policymakers more pragmatic.

Why economic evaluation prefers effectiveness evidence

Economic evaluation estimates the costs and consequences that a decision maker can expect if a technology is adopted in a particular system. An explanatory trial departs from that system in who is treated, by whom, with what adherence support and often against what comparator, so it answers a narrower question. PRECIS-2 cites widespread concern that explanatory trials can predict real-world effectiveness poorly.

The ISPOR task force on cost-effectiveness analysis alongside clinical trials recommended in 2005 that such trials be designed to evaluate effectiveness rather than efficacy, and its 2015 update repeated this with the qualification "when possible". Both reports also called for an incremental analysis on an intention-to-treat basis.

The NICE manual (PMG36, 2022, last updated October 2026) treats high-quality randomised trials against relevant comparators as the most valid evidence of relative efficacy, but lists among their limitations that results may not generalise because of the comparator, population, setting or treatment pathway. It judges external validity by whether results apply to wider patient groups and routine clinical practice.

Costs are distorted as well as effects. A review of generalisability by Sculpher and colleagues describes protocol-driven costs, meaning resource use generated by the trial itself that routine practice would not incur. Placebo comparators create a further gap, because PMG36 asks for value compared with established practice in the NHS.

Adjusting explanatory results for routine care

Explanatory trials remain valuable, since PMG36 regards randomised trials as giving a minimally biased estimate and as most appropriate for measuring relative treatment effects. Sculpher, Claxton, Drummond and McCabe argued that a single trial is usually an inadequate basis for adoption decisions and that trial evidence should be one input to a wider framework of evidence synthesis and decision analysis. Adjustments include the following.

• PMG36 describes taking the relative treatment effect from randomised trials and applying it to the baseline risk of the population of interest, which observational studies can supply, and asks analysts to check whether the relative effect is constant across baseline risks.
• The 2004 generalisability review by Sculpher and colleagues notes that compliance is generally higher in trials than in practice, and that models can give a more representative estimate by building in the reduced compliance seen in clinical practice.
• The same review describes removing protocol-driven costs by taking the costs of events from sources outside the trial while keeping the trial's efficacy result.
• PMG36 lists uses of non-randomised evidence that include assessing the generalisability of trial results, describing real-world populations and treatment patterns, and estimating resource use for economic models.
• Where the effect in the target population is uncertain, the same review suggests testing the model's results against effectiveness estimates from different studies in sensitivity analysis.

Worked example from trial efficacy to routine-care value

All numbers are hypothetical and illustrative. An explanatory trial of a preventive medicine reports a five-year event risk of 20% with placebo and 14% with treatment, a relative risk of 0.70 and a relative risk reduction of 0.30, with near-complete adherence. Treatment costs GBP 600 per person over five years.

$$ARR = p_0 \times RRR \times a$$

where $ARR$ is the absolute risk reduction, $p_0$ is the five-year event risk without treatment in the population being treated, $RRR$ is the relative risk reduction in people who take the medicine as in the trial, and $a$ is the share of people who adhere. Inside the trial, with $a$ taken as 1, $ARR = 0.20 \times 0.30 = 0.06$, the number needed to treat is about 16.7 and the cost per event avoided is GBP 10,000.

A population cohort suggests a routine-care baseline risk of 12%, and dispensing records suggest that half of patients keep taking the medicine. The analysis assumes that people who stop gain nothing and, for simplicity, that the GBP 600 cost per person started is unchanged.

$$ARR = 0.12 \times 0.30 \times 0.50 = 0.018$$

where 0.12 is the routine-care baseline risk, 0.30 the trial relative risk reduction and 0.50 the adherence share.

The cost per event avoided rises from GBP 10,000 to GBP 33,333, more than three times the trial figure, although the relative effect was never questioned. A full model would let medicine costs fall when people stop treatment, offsetting part of the rise, and would convert events into QALYs for an ICER. It would also test whether the relative effect holds at lower baseline risk, as PMG36 asks.

Explanatory trials and neighbouring ideas

Thorpe and colleagues stress that the explanatory and pragmatic labels refer to a trial's purpose and to the structure that follows from it. The pragmatic trial is the direct counterpart, asking whether a treatment works under usual conditions, the question that comparative effectiveness research puts to competing options.

Efficacy and effectiveness carry the same contrast for outcomes. Real-world evidence can help bridge them, since PMG36 allows non-randomised studies to assess the generalisability of trial results, while warning that they tend to carry a higher risk of bias from confounding.

Common misreadings of explanatory evidence

One misreading treats explanatory as a synonym for weak. Haynes points out that efficacy trials are very useful, since a treatment that fails under ideal conditions will almost certainly fail under usual ones, and most treatments do not survive this stage.

A second error is to label a whole trial with one word, when a trial can enrol a broad population yet enforce adherence tightly. The CONSORT extension also warns that more pragmatic is not automatically better, because design should follow the needs of those who will use the results.

A third error is to infer pragmatism from the analysis alone. Thorpe and colleagues note that intention-to-treat is also the norm in explanatory trials, especially those for regulatory approval, so an intention-to-treat result can still reflect ideal conditions in every other domain.

Sources

• Haynes B. Can it work? Does it work? Is it worth it? The testing of healthcare interventions is evolving. BMJ. 1999;319(7211):652. https://doi.org/10.1136/bmj.319.7211.652
• Loudon K, Treweek S, Sullivan F, Donnan P, Thorpe KE, Zwarenstein M. The PRECIS-2 tool: designing trials that are fit for purpose. BMJ. 2015;350:h2147. https://doi.org/10.1136/bmj.h2147
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 31 January 2022, last updated 13 October 2026. https://www.nice.org.uk/process/pmg36
• Ramsey S, Willke R, Briggs A, et al. Good research practices for cost-effectiveness analysis alongside clinical trials: the ISPOR RCT-CEA Task Force report. Value in Health. 2005;8(5):521-533. https://doi.org/10.1111/j.1524-4733.2005.00045.x
• Ramsey SD, Willke RJ, Glick H, et al. Cost-effectiveness analysis alongside clinical trials II. An ISPOR Good Research Practices Task Force report. Value in Health. 2015;18(2):161-172. https://doi.org/10.1016/j.jval.2015.02.001
• Schwartz D, Lellouch J. Explanatory and pragmatic attitudes in therapeutical trials. Journal of Chronic Diseases. 1967;20(8):637-648. Republished in Journal of Clinical Epidemiology. 2009;62(5):499-505. https://doi.org/10.1016/j.jclinepi.2009.01.012
• Sculpher MJ, Claxton K, Drummond M, McCabe C. Whither trial-based economic evaluation for health care decision making? Health Economics. 2006;15(7):677-687. https://doi.org/10.1002/hec.1093
• Sculpher MJ, Pang FS, Manca A, et al. Generalisability in economic evaluation studies in healthcare: a review and case studies. Health Technology Assessment. 2004;8(49). https://doi.org/10.3310/hta8490
• Thorpe KE, Zwarenstein M, Oxman AD, et al. A pragmatic-explanatory continuum indicator summary (PRECIS): a tool to help trial designers. CMAJ. 2009;180(10):E47-E57. https://doi.org/10.1503/cmaj.090523
• Zwarenstein M, Treweek S, Gagnier JJ, et al. Improving the reporting of pragmatic trials: an extension of the CONSORT statement. BMJ. 2008;337:a2390. https://doi.org/10.1136/bmj.a2390