Contents
On this page
- Definition of a trial-based economic evaluation
- Statistical power for economic endpoints
- Collection of resource-use and outcome data
- Discounting and the time horizon
- Presentation of uncertainty for decision-makers
- Missing cost and utility data
- The ISPOR good practice framework
- Boundary between within-trial analysis and decision modelling
- Key points
- Application to your study
- Frequently asked questions
Academy · Research Data Analytics & AI · Health economics in trials
Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations
An economic evaluation conducted alongside a trial inherits the trial’s design and is almost never powered for its own question. Trial power calculations are based on the primary clinical outcome rather than on cost or cost-effectiveness endpoints, because resource-use measures are highly variable; a cost-effectiveness result from a trial is therefore an estimate with wide uncertainty and should be presented as such. This does not justify omitting the evaluation, but it requires that the evaluation be designed deliberately and its uncertainty reported transparently.
Economic evaluation is frequently added to trial protocols at a late stage, by an analyst who was not involved in the design, and is then expected to answer a question the trial was not designed to answer. The predictable consequences are an unvalidated resource-use questionnaire, a time horizon shorter than the course of the condition, and a cost-effectiveness ratio reported as a point estimate without an adequate account of its uncertainty. The features that make a trial-based evaluation useful are determined at the design stage, and few of them can be added retrospectively.
Learning objectives
- Why a trial is almost never powered for its economic endpoint, and the implications for design
- The resource use to be collected, the perspective adopted, and where the measurement burden falls
- The principles of discounting and the rates applicable in different jurisdictions
- The point at which within-trial analysis must give way to decision modelling
- The requirements of the ISPOR good practice framework for trial-based cost-effectiveness analysis
- Presenting uncertainty so that a payer can use the result rather than discount it
Common misconceptions
- “If the trial is powered for the clinical outcome, the economic analysis is also adequately powered.” This is not the case. Power calculations are almost always based on the primary clinical outcome rather than on cost or cost-effectiveness endpoints, because resource-use measures are highly variable; the economic estimate is therefore typically far less precise than the clinical estimate from the same trial.
- “Cost data can be collected retrospectively from records at the end of the trial.” Resource use that was not specified prospectively is incomplete, and differentially so by arm, since the arm with more contacts generates more retrievable records. The instrument and the collection points should be specified in the protocol.
- “The cost-effectiveness ratio is the result.” A ratio with no characterisation of its uncertainty is not usable by a decision-maker. The distribution of the incremental cost and effect pair, and the probability of cost-effectiveness across a range of thresholds, are the result.
- “Discounting is a technical detail of concern only to the health economist.” Discounting alters the result whenever costs and benefits occur at different times, and the applicable rate is set by the jurisdiction — NICE recommends discounting both costs and outcomes at 3.5% annually in economic evaluations for England and Wales.
- “The trial horizon is the relevant horizon.” The relevant horizon is the period over which costs and consequences differ between the options. If the condition is chronic and the trial lasted one year, the within-trial analysis does not address the relevant decision problem, and a model is required to do so.
Definition of a trial-based economic evaluation
An economic evaluation compares two or more options on both their costs and their consequences, and reports the trade-off. Conducted alongside a trial, it uses the trial’s randomised comparison for the effect side and collects resource use within the trial for the cost side, which confers a distinctive strength and a distinctive weakness 8,21.
The strength is internal validity on the effect: randomisation means that the difference in consequences is attributable to the intervention in a way that no observational cost study can claim. The weakness is that the trial’s design decisions — its population, its protocol-driven resource use, its follow-up duration — were made for a clinical question and constrain the economic one. Protocol-mandated visits and tests inflate costs in both arms in ways that routine practice would not, and the eligibility criteria define a population narrower than the one about which a payer is deciding 14,21.
The implication is that a within-trial analysis is one input to a decision rather than the decision 20. It is the best available estimate of incremental cost and effect over the trial period in the trial population, and it should be presented explicitly as such. The Pubrica health economics and cost-effectiveness modelling pages describe how the two are related.
Statistical power for economic endpoints
This is the most important consideration in trial-based economic evaluation, and it does not reflect a deficiency in any particular trial. Trial power calculations are almost always based on the primary clinical outcome rather than on cost or cost-effectiveness endpoints, because resource-use measures are highly variable 24,25. Cost distributions are typically skewed with a long right tail: most participants consume a modest amount and a few consume a great deal, often because of a single admission. Detecting a difference in a mean that is driven by rare high-cost events requires sample sizes well beyond those needed to detect a difference in a clinical outcome.
Three consequences follow, and all three should be stated in the protocol rather than identified at the analysis stage.
First, the economic analysis should be framed as estimation rather than hypothesis testing 6,8. Reporting whether the cost difference was statistically significant is largely uninformative at trial sample sizes and invites an erroneous conclusion when a large mean difference fails to reach statistical significance.
Second, the uncertainty will be wide and must be reported; this is an accurate representation of the evidence rather than a failure of the analysis.
Third, where a decision depends on cost-effectiveness, the trial sample size should be considered with that in mind even if the trial is not formally powered for it; if the trial cannot be made sufficiently large, the plan should state explicitly that the within-trial estimate will inform a model rather than stand alone 20. Our page on imprecision describes how to interpret and rate a wide interval.
Collection of resource-use and outcome data
Four decisions must be made before the first participant is enrolled 21,27.
Perspective. The analysis must specify whose costs are counted: the health service, the health and social care system, or society including productivity losses and informal care 14,21. The perspective determines which data are collected, and a perspective chosen after collection is, in effect, chosen to fit the data. Most reimbursement authorities specify one.
Resource categories. Inpatient admissions with length of stay, outpatient and primary care contacts by type, procedures, medication with dose and duration, diagnostics, community and social care, equipment, and — if the perspective extends that far — time off work and unpaid care. For each, the source should be specified: the trial record, a routine database, a participant-completed instrument, or linkage.
Protocol-driven versus routine resource use. Every resource item that exists only because of the trial should be flagged, so that it can be excluded or handled explicitly. Failure to do so is the most common error in trial-based costing, and it biases both arms upward, usually unequally 6,14.
Preference-based outcome measure. If the analysis is a cost-utility analysis, a validated preference-based measure must be administered at pre-specified points sufficiently frequent to capture the trajectory of change, because quality-adjusted life years are computed by interpolating between measurement points. Measurement at baseline and 12 months only will miss an early benefit entirely. Further guidance is provided in the Pubrica pages on PRO instrument types and psychometric services.
Resource categories, perspective and collection source
| Resource category | Included under which perspective | Usual collection source | Where it goes wrong |
|---|---|---|---|
| Inpatient admissions and length of stay | All perspectives | Trial record, hospital database or linkage | Readmissions at non-trial hospitals are missed unless linkage is national |
| Outpatient, primary and community care contacts | All perspectives | Participant-completed instrument or routine data | Participant recall degrades quickly; contact type is often not distinguishable in routine data |
| Medication, with dose and duration | All perspectives | Trial record plus dispensing data | Over-the-counter and concomitant medicines are under-captured |
| Protocol-driven visits and tests | Excluded or handled explicitly in all perspectives | Trial schedule | Not flagged at all, inflating both arms and often unequally |
| Social care and equipment | Health and social care perspective and wider | Participant instrument, local authority or provider records | Omitted by default when the perspective was never explicitly chosen |
| Productivity loss and informal care | Societal perspective only | Participant instrument | Collected without a stated valuation method, making the figures unusable |
Structure follows standard practice for trial-based economic evaluation. The perspective is a decision to make before collection begins, since it determines which categories are in scope; many reimbursement authorities specify one.
The economic evaluation paragraph in a trial protocol (values below are illustrative)
Added late, answers nothing
A within-trial economic evaluation will be conducted to assess the cost-effectiveness of the intervention compared with usual care. Healthcare resource use will be collected during the trial and quality of life will be measured using a standard questionnaire. An incremental cost-effectiveness ratio will be calculated and reported with sensitivity analyses.
No perspective stated. The resource categories to be collected are therefore undefined, and whatever is collected will be rationalised into a perspective afterwards.
No acknowledgement of power. The paragraph implies a cost-effectiveness conclusion the trial cannot support, since power rests on the clinical outcome and resource use is highly variable.
“A standard questionnaire” is not an instrument. Nor are measurement timepoints given, and quality-adjusted life years are computed by interpolating between them.
Protocol-driven resource use unaddressed. This is the most common costing error, and it inflates both arms, usually unequally.
No discounting or horizon statement. The analysis will therefore use whatever rate the analyst chooses, which a reimbursement reviewer will recalculate.
An ICER as the output. A ratio with no joint distribution and no threshold range is not usable by a decision-maker.
No relationship to a model. If the condition is chronic, the within-trial estimate cannot answer the decision question, and the paragraph does not specify what will.
Designed in, answers something
A within-trial economic evaluation will estimate the incremental cost per quality-adjusted life year of the intervention compared with usual care over the 18-month trial period, from a health and social care perspective as specified by the reimbursement authority to which the evidence will be submitted. The trial is powered on the primary clinical outcome; because resource-use data are highly variable, the economic analysis is pre-specified as estimation rather than hypothesis testing, and the incremental cost and effect pair will be reported with its joint distribution on the cost-effectiveness plane together with the probability of cost-effectiveness across willingness-to-pay thresholds from zero to twice the authority’s reference threshold. Resource use will be collected at baseline, 6, 12 and 18 months using the resource-use instrument in Appendix D, supplemented by national hospital and dispensing data linkage so that admissions outside participating hospitals are captured; every protocol-driven visit and test is identified in Appendix E and excluded from the costing. The preference-based outcome measure will be administered at baseline and at 3, 6, 12 and 18 months, with the three-month timepoint included specifically because the intervention’s expected benefit occurs early and two-point interpolation would miss it. Costs and quality-adjusted life years will both be discounted at 3.5% annually for the period beyond 12 months. Missing costs and utilities will be imputed jointly with the clinical variables that predict them, using the method specified in the analysis plan, and complete-case and imputed analyses will be reported side by side. The within-trial result will additionally serve as the primary input to a lifetime decision-analytic model, which is described separately and is the basis of the reimbursement submission; the within-trial analysis is not presented as a lifetime estimate.
The perspective is named and justified by the audience for the evidence, which fixes the resource categories before collection starts.
The power limitation is stated and the analysis re-framed as estimation, which prevents a meaningless significance claim at analysis.
Instrument and timepoints are specified, with a reason for the early timepoint tied to when the benefit is expected.
Protocol-driven resource use is enumerated and excluded, which is the difference between a trial cost and a real-world cost.
Linkage is specified so that admissions outside participating hospitals do not silently disappear from the costing.
The discount rate is stated at the jurisdictional value and applied to outcomes as well as costs.
Uncertainty is presented as a distribution and a threshold curve, and the relationship between the within-trial analysis and the model is made explicit so neither is asked to do the other’s job.
The first paragraph commits to an output that the trial cannot produce and specifies none of the steps required during the trial to produce it. The second resolves, before enrolment, every question that cannot be answered afterwards, which comprises nearly all of them.
Discounting and the time horizon
Costs and benefits occurring at different times are not directly comparable, and discounting is the adjustment that makes them so. NICE recommends discounting both costs and outcomes at a fixed annual rate of 3.5% in economic evaluations for England and Wales 8. Two aspects of discounting are routinely mishandled.
The first is that outcomes are discounted as well as costs. Discounting only costs makes an intervention with upfront costs and delayed benefits look better than the framework intends, and it is an error rather than a conservative choice.
The second is that the rate is jurisdictional. Reimbursement authorities differ, and an evaluation prepared for one authority and submitted to another with the original rate will be rejected or recalculated. If a submission is planned to more than one jurisdiction, the base case should be presented at the rate required by the primary audience and the alternative rates as sensitivity analyses — see HTA submission and EU joint clinical assessment.
Discounting is relevant only over a horizon long enough for the timing of costs and consequences to differ, which raises the related decision of the time horizon. The time horizon should be long enough to capture all differences in cost and consequence between the options 20,21. For a single acute episode, the trial duration may suffice. For a chronic condition, a progressive disease or an intervention with late harms, it does not, and the gap between the trial horizon and the relevant horizon is precisely the gap that a model is required to fill.
Presentation of uncertainty for decision-makers
The incremental cost-effectiveness ratio is a ratio of two uncertain quantities, which is why its uncertainty cannot be summarised by a conventional confidence interval in the ordinary way: a ratio whose denominator can approach zero behaves poorly in statistical terms, and the same numerical ratio can arise from an intervention that is cheaper and less effective as from one that is more expensive and more effective 23,24.
A decision-maker can instead use the following: the joint distribution of incremental costs and incremental effects, shown on the cost-effectiveness plane so that the quadrant is visible; the net benefit at specified willingness-to-pay thresholds, which converts a ratio into a quantity with an approximately normal distribution; and the probability that the intervention is cost-effective across a range of thresholds rather than at a single one, since no payer’s threshold is a precise number 23,25.
Alongside these, deterministic sensitivity analysis should address the assumptions that are choices rather than estimates — the perspective, the discount rate, the unit costs applied, the handling of protocol-driven resource use, and the method used for missing cost data. The purpose is to show which assumption the conclusion depends on. If the conclusion reverses when a single unit cost is varied within its plausible range, that is the principal finding, and it is more informative than the base-case ratio. The Pubrica payer evidence review work focuses largely on anticipating which of these assumptions a reviewer will scrutinise.
Missing cost and utility data
Cost and utility data are missing more often than clinical outcome data and they are missing differently, which makes a default approach unsafe.
Resource-use questionnaires are long and are completed less reliably by participants who are unwell, which means that missingness correlates with both cost and outcome. Participants who withdraw tend to have incomplete cost profiles, with the most expensive component — the admission that preceded the withdrawal — sometimes present and sometimes not. A utility measure missing at a single timepoint also disrupts the interpolation on which quality-adjusted life year calculation depends, so that one gap removes a whole interval rather than one observation 6,8.
Three practical responses are recommended. The imputation approach should be specified in advance and should be multivariate, imputing costs and utilities together with the clinical variables that predict them, since imputing a cost without the outcome that drove it destroys the association being estimated 6. Complete-case and imputed analyses should be reported side by side so that a reader can see how much the imputation contributes. The assumed missingness mechanism should also be stated, with a sensitivity analysis under an alternative mechanism, as for the clinical outcome; the analysis plan is where this belongs.
The ISPOR good practice framework
The ISPOR RCT-CEA Task Force report, published in 2005, set out good research practice for cost-effectiveness analysis conducted alongside a randomised trial, and it remains the structural reference for this specific design 21. Its contribution was to treat the trial-based economic evaluation as a design problem rather than an analysis problem: the questions it asks are about the trial’s suitability for the economic question, the measurement of resource use and outcomes, the analytic approach, and the reporting.
The items of greatest practical consequence are those concerning whether the trial design supports the economic question at all — comparator relevance, population relevance, the influence of protocol-driven resource use, and the adequacy of the time horizon 20,21. A trial comparing an intervention against a comparator that is not the relevant standard of care produces a clinically interesting result but an economically uninformative one, and this cannot be corrected at the analysis stage.
For reporting, CHEERS is the applicable checklist for economic evaluations and should be used alongside CONSORT 2025 for the trial itself. Presenting the economic evaluation as a few paragraphs within the clinical paper is common practice and a common reason that the evaluation cannot be assessed or reused.
Boundary between within-trial analysis and decision modelling
Four situations require a decision-analytic model rather than a within-trial analysis, and recognising them early avoids considerable wasted effort.
The horizon is shorter than the condition. A one-year trial in a chronic disease cannot capture the lifetime cost and consequence difference, so the within-trial result is an input to a model that extrapolates it, with the extrapolation assumptions stated and tested 20.
The comparator set is incomplete. A trial evaluates only the comparators it includes. A decision usually involves several options, and comparing all of them requires synthesis across trials — see network meta-analysis.
The trial population is narrower than the decision population. A model can be populated with subgroup-relevant parameters in a way a single trial cannot.
Intermediate endpoints need linking to final outcomes. If the trial measured a surrogate, the step from surrogate to survival or quality-adjusted life years is a modelled relationship, and its uncertainty has to be carried through rather than assumed away.
In all four, the within-trial analysis remains worthwhile: it is the most internally valid estimate available of the incremental cost and effect over the observed period, and it anchors the model. It should not, however, be presented as the answer to a question it cannot address. Budget impact modelling and value dossier work pick up from there.
What the within-trial analysis can and cannot deliver
| Question | Within-trial analysis | What is needed instead |
|---|---|---|
| Incremental cost and effect over the trial period, in the trial population | Yes — this is its strength, with randomisation protecting the effect estimate | — |
| A precise cost difference | Rarely — power is based on the clinical outcome and resource use is highly variable | Wide interval reported transparently; larger sample if the decision depends on it |
| Lifetime cost-effectiveness in a chronic condition | No — the horizon is too short | A decision-analytic model with stated extrapolation assumptions |
| Comparison across all relevant treatment options | No — the trial evaluates only the comparators it includes | Evidence synthesis, typically network meta-analysis, feeding a model |
| Cost-effectiveness in a population wider than the eligibility criteria | No | A model populated with parameters for the decision population |
| Final outcomes when only a surrogate was measured | No | A modelled surrogate-to-outcome relationship with its uncertainty carried through |
The distinction matters for how the analysis is presented: a within-trial result offered as the answer to a lifetime question will be rejected by a reimbursement reviewer, while the same result offered as a model input is valuable.
Key points
Key points
A trial-based economic evaluation is strong on the effect side, because randomisation protects it, and weak on precision, because trial power calculations are based on the primary clinical outcome rather than on cost or cost-effectiveness endpoints and resource-use measures are highly variable. The design should reflect this: the perspective should be specified before any data are collected, the resource-use instrument and its collection points defined, protocol-driven resource use identified and excluded, and the preference-based measure administered often enough that interpolation does not miss the benefit. Costs and outcomes should both be discounted at the jurisdictional rate — NICE recommends 3.5% annually for England and Wales — and the analysis framed as estimation, reporting the joint distribution of incremental cost and effect and the probability of cost-effectiveness across a range of thresholds rather than a single ratio. The ISPOR RCT-CEA Task Force report remains the structural reference for the design. Where the trial horizon is shorter than the course of the condition, this should be recognised early, because the within-trial result then becomes a model input rather than an answer.
Application to your study
Application to your study
- The analytical perspective should be selected first. The options are the health service, health and social care, or societal perspective; the choice should reflect the authority to which the evidence will be submitted and should be made before any resource-use data are collected.
- The protocol should state that the trial is not powered for the economic endpoint. The economic analysis should then be pre-specified as estimation rather than hypothesis testing.
- Protocol-driven resource use should be enumerated in an appendix. This includes every visit, test and contact that exists only because of the trial, so that it can be explicitly excluded from the costing.
- Utility measurement points should be set according to when the benefit is expected. Quality-adjusted life years are interpolated between measurements, so a benefit that peaks between two distant points will not appear at all.
- Data linkage should be specified for resource use that the trial record does not capture. Admissions and prescriptions outside participating centres are not visible in the trial record and are systematically absent from the costing unless linkage is used.
- Costs and outcomes should be discounted at the jurisdictional rate. The rate is 3.5% annually for NICE evaluations in England and Wales, with other rates included as sensitivity analyses if a second submission is planned.
- The distribution, rather than the ratio alone, should be reported. This includes the cost-effectiveness plane, net benefit at stated thresholds and the probability of cost-effectiveness across a range of thresholds.
- The need for a decision model should be determined at the design stage. If the horizon is shorter than the condition, the protocol should state this and position the within-trial analysis as the primary input to the model.
Frequently asked questions
Trials are rarely powered for these endpoints because power calculations are almost always based on the primary clinical outcome, and resource-use measures are highly variable. Cost distributions in trials are heavily skewed: most participants consume a modest amount and a small number consume a great deal, usually because of one admission. A mean difference driven by rare high-cost events requires a much larger sample to estimate precisely than a difference in a clinical outcome does. Two consequences follow. The economic analysis should be pre-specified as estimation rather than hypothesis testing, since a significance test on cost at trial sample sizes is close to uninformative. In addition, the resulting uncertainty will be wide, which accurately reflects the evidence rather than indicating an analytic failure.
The rate specified by the relevant reimbursement authority, applied to both costs and outcomes. For evaluations in England and Wales, NICE recommends discounting both at a fixed annual rate of 3.5%. Two errors are common. The first is discounting costs but not outcomes, which systematically favours interventions with upfront costs and delayed benefits and is an error rather than a conservative choice. The second is carrying one jurisdiction’s rate into another’s submission, which will be recalculated or rejected. Where submission in more than one jurisdiction is expected, the base case should be presented at the primary audience’s required rate and the alternatives as sensitivity analyses, and the rate should be stated explicitly wherever the results appear.
No. Failure to handle it is the most common costing error in trial-based evaluations. Protocol-mandated visits, tests and monitoring exist because of the trial, not because of the condition or the intervention, and including them inflates the cost of both arms. Moreover, the inflation is usually unequal: the arm requiring more monitoring accrues more protocol-driven cost, which produces a cost difference that would not exist in routine practice. The recommended approach is to enumerate every protocol-driven item in a protocol appendix at the design stage, cost the arms with and without them, and report the routine-practice costing as the base case with the full trial costing as a sensitivity analysis.
Jointly, with the clinical variables that predict them, under a stated missingness assumption. Cost and utility data are missing more often than clinical outcomes, and the missingness is informative: long resource-use questionnaires are completed less reliably by participants who are unwell, and withdrawal is often preceded by precisely the high-cost event that most needs to be captured. Imputing a cost without the outcome that drove it destroys the association being estimated, so the imputation model should be multivariate. Complete-case and imputed analyses should be reported side by side so that a reader can see how much work the imputation is doing, and a sensitivity analysis under a different missingness assumption should be included, as for the clinical outcome.
A within-trial analysis estimates incremental cost and effect over the period observed, in the population enrolled, using the trial’s own data. Its strength is internal validity: randomisation protects the effect estimate. A decision-analytic model extrapolates beyond the trial — to a longer horizon, to a wider population, to a fuller comparator set, or from a surrogate to a final outcome — using the trial result as one parameter among several. Four situations require a model: the horizon is shorter than the condition, the comparator set is incomplete, the trial population is narrower than the decision population, or only an intermediate endpoint was measured. In all four the within-trial analysis is still worth doing; it anchors the model rather than answering the decision.
As a distribution rather than a single number. An incremental cost-effectiveness ratio is a ratio of two uncertain quantities whose denominator can approach zero, so it behaves poorly in statistical terms, and the same ratio value can arise from opposite situations — cheaper and less effective, or more expensive and more effective. A decision-maker can use the joint distribution of incremental costs and effects plotted on the cost-effectiveness plane, so that the quadrant is visible; net benefit at specified willingness-to-pay thresholds, which converts the ratio into a better-behaved quantity; and the probability of cost-effectiveness across a range of thresholds rather than at one, since no payer’s threshold is an exact number.
The ISPOR RCT-CEA Task Force report, published in 2005, set out good research practice specifically for cost-effectiveness analysis conducted alongside a randomised trial, and it remains the structural reference for this design. Its principal contribution is treating the evaluation as a design problem: it asks whether the trial’s comparator, population, protocol-driven resource use and time horizon can support an economic question at all, before any analysis questions arise. For reporting, CHEERS is the applicable checklist for economic evaluations and should be used alongside CONSORT for the trial report itself. Presenting the economic evaluation as a few paragraphs within the clinical paper is both common practice and the main reason such evaluations cannot be assessed or reused.
At the design stage, before the protocol is drafted. Almost every decision that determines whether an economic evaluation will be usable cannot be made retrospectively: the perspective, the resource-use instrument and its collection points, the identification of protocol-driven resource use, the choice and frequency of the preference-based outcome measure, the linkage arrangements for care received outside participating centres, the time horizon, and whether the comparator is the relevant standard of care. A health economist engaged at the analysis stage can compute a ratio from whatever was collected, and that is usually the extent of what is available — which is why economic evaluations added late so often produce results that are seldom used.
References
- Rogers A, De Paoli G, Subbarayan S, Copland R, Harwood K, Coyle J, et al. A Systematic Review of Methods used to Conduct Decentralised Clinical Trials. British Journal of Clinical Pharmacology. 2022;88(6):2843-2862. doi:10.1111/bcp.15205 read 9 October 2026. doi.org/10.1111/bcp.15205
- Zeng L, Brignardello-Petersen R, Hultcrantz M, Mustafa RA, Murad MH, Iorio A, et al. GRADE Guidance 34: update on rating imprecision using a minimally contextualized approach. Journal of Clinical Epidemiology. 2022;150:216-224. doi:10.1016/j.jclinepi.2022.07.014 read 9 October 2026. doi.org/10.1016/j.jclinepi.2022.07.014
- Schunemann HJ, Neumann I, Hultcrantz M, Brignardello-Petersen R, Zeng L, Murad MH, et al. GRADE guidance 35: update on rating imprecision for assessing contextualized certainty of evidence and making decisions. Journal of Clinical Epidemiology. 2022;150:225-242. doi:10.1016/j.jclinepi.2022.07.015 read 9 October 2026. doi.org/10.1016/j.jclinepi.2022.07.015
- de Jong AJ, Grupstra RJ, Santa-Ana-Tellez Y, Zuidgeest MGP, de Boer A, Gardarsdottir H. Which decentralised trial activities are reported in clinical trial protocols of drug trials initiated in 2019-2020? A cross-sectional study in ClinicalTrials.gov. BMJ Open. 2022;12(8):e063236. doi:10.1136/bmjopen-2022-063236 read 9 October 2026. doi.org/10.1136/bmjopen-2022-063236
- DeVito NJ, Bacon S, Goldacre B. Compliance with legal requirement to report clinical trial results on ClinicalTrials.gov: a cohort study. The Lancet. 2020;395(10221):361-369. doi:10.1016/S0140-6736(19)33220-9 read 9 October 2026. doi.org/10.1016/S0140-6736(19)33220-9
- Hughes D, Charles J, Dawoud D, Edwards RT, Holmes E, Jones C, et al. Conducting Economic Evaluations Alongside Randomised Trials: Current Methodological Issues and Novel Approaches. PharmacoEconomics. 2016;34(5):447-461. doi:10.1007/s40273-015-0371-y read 9 October 2026. doi.org/10.1007/s40273-015-0371-y
- Anderson ML, Chiswell K, Peterson ED, Tasneem A, Topping J, Califf RM. Compliance with results reporting at ClinicalTrials.gov. New England Journal of Medicine. 2015;372(11):1031-1039. doi:10.1056/NEJMsa1409364 read 9 October 2026. doi.org/10.1056/NEJMsa1409364
- Petrou S, Gray A. Economic evaluation alongside randomised controlled trials: design, conduct, analysis, and reporting. BMJ. 2011;342:d1548. doi:10.1136/bmj.d1548 read 9 October 2026. doi.org/10.1136/bmj.d1548
- Guyatt GH, Oxman AD, Kunz R, Brozek J, Alonso-Coello P, Rind D, et al. GRADE guidelines 6. Rating the quality of evidence - imprecision. Journal of Clinical Epidemiology. 2011;64(12):1283-1293. doi:10.1016/j.jclinepi.2011.01.012 read 9 October 2026. doi.org/10.1016/j.jclinepi.2011.01.012
- Guyatt G, Oxman AD, Akl EA, Kunz R, Vist G, Brozek J, et al. GRADE guidelines: 1. Introduction - GRADE evidence profiles and summary of findings tables. Journal of Clinical Epidemiology. 2011;64(4):383-394. doi:10.1016/j.jclinepi.2010.04.026 read 9 October 2026. doi.org/10.1016/j.jclinepi.2010.04.026
- Balshem H, Helfand M, Schunemann HJ, Oxman AD, Kunz R, Brozek J, et al. GRADE guidelines: 3. Rating the quality of evidence. Journal of Clinical Epidemiology. 2011;64(4):401-406. doi:10.1016/j.jclinepi.2010.07.015 read 9 October 2026. doi.org/10.1016/j.jclinepi.2010.07.015
- Schulz KF, Altman DG, Moher D. CONSORT 2010 Statement: updated guidelines for reporting parallel group randomised trials. Trials. 2010;11(1):32. doi:10.1186/1745-6215-11-32 read 9 October 2026. doi.org/10.1186/1745-6215-11-32
- Bassler D, Briel M, Montori VM, Lane M, Glasziou P, Zhou Q, et al. Stopping randomized trials early for benefit and estimation of treatment effects: systematic review and meta-regression analysis. JAMA. 2010;303:1180-1187. read 9 October 2026. school16.sipta.org
- Polsky D, Glick H. Costing and cost-analysis in randomized controlled trials. PharmacoEconomics. 2009;27:179-188. read 9 October 2026. ouci.dntb.gov.ua
- Guyatt GH, Oxman AD, Vist GE, Kunz R, Falck-Ytter Y, Alonso-Coello P, et al. GRADE: an emerging consensus on rating quality of evidence and strength of recommendations. BMJ. 2008;336:924-926. read 9 October 2026. cobe.paginas.ufsc.br
- Schunemann HJ, Oxman AD, Brozek J, Glasziou P, Jaeschke R, Vist GE, et al. Grading quality of evidence and strength of recommendations for diagnostic tests and strategies. BMJ. 2008;336:1106-1110. read 9 October 2026. cobe.paginas.ufsc.br
- Guyatt GH, Oxman AD, Kunz R, Vist GE, Falck-Ytter Y, Schunemann HJ. What is "quality of evidence" and why is it important to clinicians? BMJ. 2008;336:995-998. read 9 October 2026. cobe.paginas.ufsc.br
- von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP; STROBE Initiative. Strengthening the reporting of observational studies in epidemiology (STROBE) statement: guidelines for reporting observational studies. BMJ. 2007;335(7624):806-808. doi:10.1136/bmj.39335.541782.AD read 9 October 2026. doi.org/10.1136/bmj.39335.541782.AD
- von Elm E, Altman DG, Egger M, Pocock SJ, Gotzsche PC, Vandenbroucke JP; STROBE Initiative. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Epidemiology. 2007;18(6):800-804. doi:10.1097/EDE.0b013e3181577654 read 9 October 2026. doi.org/10.1097/EDE.0b013e3181577654
- Sculpher MJ, Claxton K, Drummond M, McCabe C. Wither trial-based economic evaluation for health care decision making? Health Economics. 2006;15:677-687. read 9 October 2026. ouci.dntb.gov.ua
- Ramsey S, Willke R, Briggs A, Brown R, Buxton M, Chawla A, Cook J, Glick H, Liljas B, Petitti D, Reed S. 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. doi:10.1111/j.1524-4733.2005.00045.x read 9 October 2026. doi.org/10.1111/j.1524-4733.2005.00045.x
- Montori VM, Devereaux PJ, Adhikari NK, Burns KEA, Eggert CH, Briel M, et al. Randomized trials stopped early for benefit: a systematic review. JAMA. 2005;294:2203-2209. read 9 October 2026. school16.sipta.org
- Willan AR. Analysis, sample size, and power for estimating incremental net health benefit from clinical trial data. Controlled Clinical Trials. 2001;22:228-237. read 9 October 2026. ouci.dntb.gov.ua
- Willan AR, O'Brien BJ. Sample size and power issues in estimating incremental cost-effectiveness ratios from clinical trials data. Health Economics. 1999;8:203-211. read 9 October 2026. ouci.dntb.gov.ua
- Briggs AH, Gray AM. Power and sample size calculations for stochastic cost-effectiveness analysis. Medical Decision Making. 1998;18(Suppl):S81-S92. read 9 October 2026. ouci.dntb.gov.ua
- Pogue JM, Yusuf S. Cumulating evidence from randomized trials: utilizing sequential monitoring boundaries for cumulative meta-analysis. Controlled Clinical Trials. 1997;18:580-593. read 9 October 2026. school16.sipta.org
- Glasziou PP, Simes RJ, Hall J. Design of a cost-effectiveness study within a randomized trial: the LIPID trial for secondary prevention of IHD. Controlled Clinical Trials. 1997;18:464-476. read 9 October 2026. ouci.dntb.gov.ua
- Pocock SJ, Hughes MD. Practical problems in interim analyses, with particular regard to estimation. Controlled Clinical Trials. 1989;10:209S-221S. read 9 October 2026. school16.sipta.org
Every source above is a peer-reviewed article, a publisher’s record of one, or a standard issued by the body that owns it. Each carries the date it was last read. Where a standard moves, this article is revised rather than left up with a note.
How to cite this article
Pick your referencing style and copy it. Your university or target journal may vary the punctuation — check its guide if it has one.
Pubrica Academy. (2026, October 9). Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations. Pubrica. https://pubrica.com/academy/research/health-economics-in-clinical-trials/
Pubrica Academy. “Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations.” Pubrica, 9 Oct. 2026, pubrica.com/academy/research/health-economics-in-clinical-trials/.
Pubrica Academy (2026) Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations. Pubrica. Available at: https://pubrica.com/academy/research/health-economics-in-clinical-trials/ (Accessed: 15 October 2026).
Pubrica Academy. “Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations.” Pubrica. October 9, 2026. https://pubrica.com/academy/research/health-economics-in-clinical-trials/
Pubrica Academy, ‘Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations’ (Pubrica, 9 October 2026) <https://pubrica.com/academy/research/health-economics-in-clinical-trials/> accessed 15 October 2026.
Pubrica Academy. Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations. [Internet]. Pubrica; 2026 Oct 9 [cited 2026 Oct 15]. Available from: https://pubrica.com/academy/research/health-economics-in-clinical-trials/
Pubrica Academy, “Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations,” Pubrica, Oct. 9, 2026. [Online]. Available: https://pubrica.com/academy/research/health-economics-in-clinical-trials/
@misc{health_economics_in_clinical_trials_2026,
author = {{Pubrica Academy}},
title = {Health Economics in Clinical Trials: Design, Analysis and Reporting of Trial-Based Economic Evaluations},
howpublished = {Pubrica},
year = {2026},
month = {oct},
url = {https://pubrica.com/academy/research/health-economics-in-clinical-trials/},
note = {Accessed: <span data-cite-acc="iso">2026-10-15</span>}
}This article is written to be learned from. Cite it if you are referring to what it says; if you are using the evidence behind it, cite the sources listed above instead.
Trial-based economic evaluation design checklist
This checklist should be completed during protocol development. Almost nothing on it can be decided later: by the time data collection has begun, most of these questions have been settled by default.
TRIAL-BASED ECONOMIC EVALUATION DESIGN CHECKLIST
Trial: ______________________ Protocol version: ______________
Health economist involved from: [ ] design [ ] later - when: ___
1. THE DECISION THIS EVIDENCE SERVES
Audience / reimbursement authority: ________________________
Decision to be informed: ___________________________________
Required perspective for that audience: ____________________
Required discount rate: ______% applied to costs AND outcomes?
[ ] yes
Reference threshold used by that audience: _________________
2. ANALYSIS TYPE
[ ] Cost-effectiveness (natural units) unit: ______________
[ ] Cost-utility (QALYs)
[ ] Cost-consequence
[ ] Cost-minimisation - justify equivalence assumption: ____
3. POWER AND FRAMING (state this in the protocol)
Trial powered on: __________________________________________
[ ] Protocol states the trial is NOT powered on the economic
endpoint
[ ] Economic analysis pre-specified as ESTIMATION, not testing
Expected direction of the cost difference, and why: ________
4. TIME HORIZON
Trial follow-up: ______ months
Period over which costs/consequences actually differ: ______
[ ] Trial horizon is adequate
[ ] Model required - described in: ________________________
5. RESOURCE USE
Instrument / source for each category:
Inpatient + length of stay: ______________________________
Outpatient / primary / community: ________________________
Medication (dose, duration): _____________________________
Diagnostics and procedures: ______________________________
Social care / equipment: _________________________________
Productivity / informal care (societal only): ____________
Collection timepoints: _____________________________________
Routine data linkage arranged? [ ] yes - which datasets: __
[ ] no - what is lost: ____
PROTOCOL-DRIVEN ITEMS enumerated in appendix? [ ] yes
Handling: [ ] excluded from base case
[ ] included, reported as sensitivity analysis
6. OUTCOME MEASUREMENT
Preference-based instrument: _______________________________
Value set / tariff to be applied: __________________________
Measurement points: _______________________________________
[ ] Points are frequent enough to capture the expected
shape of benefit (state when benefit is expected: ____)
7. UNIT COSTS
Source and price year: _____________________________________
Inflation adjustment method: _______________________________
Costs to be varied in sensitivity analysis: ________________
8. MISSING DATA
Expected completeness of resource use: ______%
Imputation approach: _______________________________________
[ ] Costs and utilities imputed JOINTLY with clinical predictors
[ ] Complete-case and imputed analyses both reported
Assumed mechanism: ______ sensitivity under alternative: ___
9. REPORTING OF UNCERTAINTY
[ ] Cost-effectiveness plane with joint distribution
[ ] Net benefit at stated thresholds
[ ] Probability of cost-effectiveness across a threshold RANGE
[ ] Deterministic sensitivity analysis on assumptions, not
just parameters
Assumption the conclusion is most sensitive to: ____________
10. REPORTING STANDARD
[ ] CHEERS mapped during drafting
[ ] CONSORT for the trial report
[ ] Economic evaluation reported in its own paper, not only
as paragraphs in the clinical paper Pubrica AcademyWritten and maintained by Pubrica’s subject matter experts. Articles are revised when the underlying standards change, and each article states the date on which it was last reviewed.
How Pubrica can help
Pubrica designs the economic evaluation while the protocol remains open, which is the only stage at which the perspective, the resource-use instrument, the utility measurement schedule and the treatment of protocol-driven resource use can be decided rather than inherited. The analysis is written as estimation, with uncertainty presented on the cost-effectiveness plane and across a range of thresholds, and the protocol states where the within-trial result ends and a model must begin. For submissions, the evaluation is mapped against CHEERS and against the specific requirements of the receiving authority.