Cost-Minimisation Analysis

This page explains the evidence required before a costs-only comparison is appropriate, why a non-significant result does not prove equivalence, and how total and incremental costs should be calculated. It also explains how perspective, time horizon, uncertainty and health technology assessment affect interpretation of the least-cost result.

Cost-minimisation analysis is commonly shortened to CMA. Although its calculation can appear simpler than other forms of economic evaluation, it has a demanding evidence requirement. Analysts must establish that plausible differences in effectiveness, safety, quality of life and other relevant outcomes would not change the decision before comparing costs alone.

What cost-minimisation analysis does

CMA identifies the least costly alternative when relevant outcomes have already been shown not to differ meaningfully. It does not itself test or prove that the alternatives are equivalent. Its conclusion remains conditional on the evidence supporting that claim.

CMA therefore differs from cost-effectiveness analysis and cost-utility analysis. Those methods retain differences in outcomes and relate them to differences in costs. CMA removes outcomes from the economic comparison only after a separate evidence assessment establishes that meaningful outcome differences are not expected to affect the decision.

The evidence gate that must be passed first

CMA is appropriate only after a defined evidence gate has been satisfied. This gate protects the analysis from treating an absence of evidence as evidence of equivalence.

1. Identify all decision-relevant outcomes. Include effectiveness, safety, adverse events, adherence, convenience, quality of life and downstream consequences where they could affect the choice.
2. Define an acceptable difference prospectively. State the clinical equivalence or non-inferiority margin and justify why differences within it would not change the decision.
3. Use suitable comparative evidence. Prefer appropriately designed and analysed equivalence or non-inferiority evidence rather than a non-significant superiority test.
4. Assess the complete evidence pattern. Examine estimates, confidence intervals, protocol deviations, missing data, adherence and consistency across relevant outcomes.
5. Confirm decision equivalence. Establish that any plausible outcome differences are too small to require explicit valuation in the economic analysis.

If this gate fails, cost-effectiveness analysis, cost-utility analysis, cost-consequence analysis or another method that keeps outcome differences visible is needed.

Which choices define a cost-minimisation analysis

A valid CMA requires both an evidence decision and a cost comparison. These choices determine whether the method is appropriate and whether its result answers the intended healthcare question.

1. Define the decision problem. Specify the population, setting, options, comparator, perspective and decision context.
2. Identify relevant outcomes. Include every consequence capable of altering the decision.
3. Specify the equivalence requirement. State the margin, endpoints, follow-up period and evidential standard.
4. Evaluate the clinical evidence. Assess design, precision, analysis population, missing data and applicability.
5. Decide whether CMA is valid. Stop if meaningful outcome differences remain plausible.
6. Define the costing scope. Identify all resources relevant to the perspective and time horizon.
7. Measure resource quantities. Record acquisition, administration, monitoring, adverse-event management, follow-up and downstream use.
8. Value resources. Apply appropriate unit costs and a stated price year.
9. Calculate total and incremental costs. Keep quantities, unit costs and subtotals auditable.
10. Test uncertainty. Examine both the equivalence assumption and the cost conclusion.
11. Interpret the decision. Prefer the least-cost option only while the evidence gate remains satisfied.

Equivalence, non-inferiority and no significant difference

These findings do not establish the same thing. The study design, prespecified margin, confidence interval and decision context determine whether the evidence can support a costs-only comparison.

Equivalence within a justified two-sided margin

• What it establishes: The confidence interval for the difference lies entirely within a range considered clinically unimportant.
• Whether it supports CMA: Potentially, provided that all relevant outcomes and other validity conditions are covered.

Non-inferiority within a justified one-sided margin

• What it establishes: The new option is not worse than the comparator by more than the accepted loss.
• Whether it supports CMA: Potentially, provided that non-inferiority is shown for all important outcomes, including safety, as HTA bodies typically require. Non-inferiority on a single endpoint is not enough, and possible gains still matter when the new option costs more.

No statistically significant difference in a superiority test

• What it establishes: The study did not detect a difference at its chosen significance level.
• Whether it supports CMA: No. Failure to detect a difference does not demonstrate equivalence.

An underpowered study can produce a wide confidence interval that contains both important benefit and important harm. Calling such treatments equivalent would conceal decision-relevant uncertainty.

Checking whether the evidence is sufficient

The evidence assessment should be explicit enough for another reader to understand why costs are being compared alone. It should address the study design, margin, precision, relevant outcomes and applicability.

• Outcomes: All material consequences must have been considered.
• Margin: The equivalence or non-inferiority margin must be clinically justified and specified independently of the observed result.
• Precision: Confidence intervals must exclude differences that would change the decision.
• Population and setting: The evidence must apply to the intended population, setting, use and comparator.
• Safety and tolerability: Relevant differences must be absent or represented explicitly.
• Perspective: The costs counted must match the stated decision-maker and analytical perspective.
• Time horizon: The analysis must capture material resource and outcome differences.
• Uncertainty: Plausible assumptions must not invalidate equivalence or reverse the cost interpretation without being reported.
• Analysis population: The analysis must use appropriate intention-to-treat and per-protocol evidence where required by the study design.
• Missing data and adherence: Missing observations, treatment switching, protocol deviations and adherence differences must not conceal meaningful outcome differences.

Measuring the full cost of each alternative

CMA should compare all material resource differences included under the stated perspective. Comparing acquisition prices alone can produce the wrong least-cost option when administration, monitoring, adverse events, follow-up or downstream care differ.

For option j:

Total Costⱼ = Σ(Resource Quantityᵢⱼ × Unit Costᵢ)

For Option A compared with Option B:

Incremental Cost = Total CostA − Total CostB

Resource quantities and unit costs should be reported separately where possible. This makes the calculations auditable and allows readers to determine whether important resources have been omitted or counted twice.

Relevant resource categories may include:

• Treatment acquisition.
• Administration.
• Diagnostic testing.
• Routine monitoring.
• Staff time.
• Consumables and equipment.
• Adverse-event management.
• Hospital admissions.
• Follow-up care.
• Patient travel and time when included by the perspective.
• Caregiver time when included by the perspective.
• Implementation and training.
• Downstream healthcare use.

Each unit cost should have a source, currency, price year and applicable setting. Any inflation adjustments, currency conversions, missing values, extrapolations or assumptions should be explained.

Worked example: choosing the least-cost treatment

Two treatments have reliable evidence of equivalent relevant outcomes and comparable safety. The following illustrative costs are per patient.

Treatment A

• Acquisition: £1,200
• Administration: £180
• Monitoring: £140
• Adverse-event management: £80
• Total cost: £1,600

Treatment B

• Acquisition: £950
• Administration: £320
• Monitoring: £160
• Adverse-event management: £90
• Total cost: £1,520

Incremental Cost = £1,600 − £1,520 = £80

Treatment B costs £80 less per patient and is the cost-minimising choice, conditional on the equivalence evidence remaining credible. If a meaningful effectiveness, safety or quality-of-life difference becomes plausible, the CMA conclusion no longer applies.

The example also shows why acquisition price should not be interpreted as total cost. Treatment B's acquisition price is £250 lower, but £170 of that saving is offset by higher administration, monitoring and adverse-event costs. All material resource differences must be included before identifying the least-cost option.

How perspective can change the least-cost option

The least-cost option is perspective-specific. A payer saving may be accompanied by additional patient travel, caregiver time, provider workload or costs elsewhere in the health system.

The analysis should state whether it uses a healthcare-system, payer, provider or societal perspective. Material excluded or shifted costs should be disclosed rather than describing one option as universally cheaper.

How time horizon and discounting affect costs

The time horizon should be long enough to capture all material cost differences and any evidence that could challenge outcome equivalence. A short horizon can miss later monitoring, adverse-event, follow-up, replacement or downstream costs.

Discounting is applied when costs occur over more than one year, in line with the relevant jurisdiction's guidance. The currency, price year, time horizon, discount rate and timing convention should be stated, and alternative assumptions tested when the conclusion is sensitive to these choices.

Testing clinical and cost uncertainty separately

Clinical and cost uncertainty each need explicit examination. Where patient-level data on both costs and outcomes are available, separate and sequential tests on outcomes and then costs can misstate decision uncertainty, so the joint distribution of cost and outcome differences should be estimated, as in a cost-effectiveness analysis. A stable cost saving cannot rescue an invalid equivalence assumption, and credible equivalence does not make an uncertain cost difference definitive.

The method's standing has been contested for more than two decades. Briggs and O'Brien argued in 2001 that separate, sequential hypothesis tests on differences in effects and costs are an inappropriate basis for choosing CMA, and that analysis should focus on the joint distribution of cost and effect differences. Dakin and Wordsworth later concluded that CMA biases measures of uncertainty and that cost-effectiveness analysis evaluating the joint distribution is almost always required to avoid biased estimation of uncertainty in trial-based economic evaluation. A NICE Decision Support Unit report by Wailoo and Dixon found that the view that CMA is inappropriate in most situations had strengthened over the preceding two decades, while setting out conditions for its use: an a priori case that similar outcomes are scientifically plausible, and non-inferiority for all important outcomes, not just the primary outcome.

Clinical-equivalence uncertainty

Clinical-equivalence uncertainty may arise from:

• The selected equivalence or non-inferiority margin.
• Endpoint selection.
• Confidence-interval width.
• Study power.
• Analysis population.
• Missing data.
• Adherence.
• Treatment switching.
• Extrapolation.
• Applicability to the intended population or setting.

Cost uncertainty

Cost uncertainty may arise from:

• Resource-use quantities.
• Unit costs.
• Adverse-event management.
• Administration requirements.
• Monitoring frequency.
• Implementation costs.
• Time horizon.
• Price year.
• Discount rate.
• Analytical perspective.

The analysis should test whether plausible changes alter the identity of the least-cost option. If they do, the circumstances under which each option becomes less costly should be reported instead of presenting one answer as certain.

When another evaluation method is needed

CMA should stop when a meaningful difference in outcomes remains plausible. The appropriate alternative method depends on how those outcomes should be represented, and switching methods protects the decision from losing important consequences.

• If equivalence is credible and incremental cost is negative, Option A is less costly.
• If equivalence is credible and incremental cost is positive, Option B is less costly.
• If equivalence is credible but the cost difference is uncertain or immaterial, that uncertainty should be reported.
• If equivalence is not credible, a CMA-based least-cost recommendation should not be issued.
• If outcomes differ in a common natural unit, cost-effectiveness analysis may be appropriate.
• If outcomes are measured using QALYs, cost-utility analysis may be appropriate.
• If several different outcomes should remain visible, cost-consequence analysis may be appropriate.

How CMA contributes to health technology assessment

CMA can provide conditional evidence about which clinically equivalent alternative has the lowest cost. Health technology assessment may use that evidence alongside clinical effectiveness, safety, budget impact, equity, feasibility and implementation requirements.

A least-cost result does not automatically establish affordability, feasibility or suitability for adoption. Budget impact analysis may still be needed to estimate the total financial consequences for a particular budget holder and planning period.

Some HTA bodies provide a formal route for costs-only comparisons. In England, the NICE methods manual (PMG36) sets out a cost-comparison analysis in technology appraisals for technologies likely to provide similar or greater health benefits at similar or lower cost than comparators recommended in published NICE guidance for the same population. In Australia, the Pharmaceutical Benefits Advisory Committee guidelines set out a cost-minimisation approach for medicines shown to be non-inferior or superior to their main comparator in both effectiveness and safety.

Implementing the calculation in Excel

A transparent spreadsheet should keep evidence inputs, resource quantities, unit costs, calculations and the final evidence gate in separate labelled fields. The least-cost recommendation should not appear unless the evidence requirement is satisfied.

• Cost for one resource: =Quantity*UnitCost
• Total cost for an option: =SUMPRODUCT(QuantityRange,UnitCostRange)
• Incremental cost: =TotalCostA-TotalCostB
• Check a two-sided equivalence interval: =IF(AND(CILower>-Margin,CIUpper<Margin),"Within margin","CMA not supported")
• Protect the decision with an evidence gate: =IF(EquivalenceConfirmed="Yes",IF(TotalCostA<TotalCostB,"Option A",IF(TotalCostA>TotalCostB,"Option B","Equal cost")),"CMA not appropriate")
• Test a cost scenario: =BaseCost*(1+ScenarioChange)

The equivalence decision, its source and its justified margin should be recorded as explicit inputs. Equivalence should not be inferred automatically from a p-value or from a blank evidence field.

What CMA can and cannot decide

CMA answers a deliberately narrow question: which option costs less when meaningful outcome differences have been ruled out. It cannot replace the clinical evidence required to justify that question or make the final adoption decision.

CMA can:

• Compare the complete relevant costs of alternatives.
• Identify the least-cost option after the evidence gate is satisfied.
• Show which resource categories drive cost differences.
• Test whether cost uncertainty changes the least-cost option.
• Support a broader health technology assessment.

CMA cannot:

• Prove that healthcare alternatives are equivalent.
• Convert a non-significant superiority test into evidence of equivalence.
• Ignore meaningful differences in effectiveness, safety or quality of life.
• Determine affordability from cost per patient alone.
• Resolve equity, feasibility or implementation questions.
• Make the final adoption or reimbursement decision.

Common mistakes and how to avoid them

CMA is particularly vulnerable to misuse because a simple cost comparison can conceal a weak equivalence claim. The most important safeguard is to make the evidence gate visible and testable.

• No statistically significant difference is not proof of equivalence.
• The equivalence margin must be clinically justified, not chosen after seeing the data.
• Non-inferiority on one endpoint does not establish equivalence across all relevant outcomes.
• An underpowered study does not establish that treatments have the same outcomes.
• Acquisition price alone is not total cost.
• Resource quantities and unit costs should not be hidden inside unexplained totals.
• The conclusion depends on perspective and time horizon.
• Clinical and cost uncertainty each require explicit assessment; sequential significance tests on outcomes and then costs are not a substitute for analysing them jointly.
• A least-cost statement is conditional on the evidence gate.
• Meaningful outcome differences require an evaluation method that represents them explicitly.

What should be reported

A report should state the decision problem, alternatives, comparator, population, setting, perspective, time horizon, currency and price year, discounting, relevant outcomes, endpoint definitions, equivalence or non-inferiority margin and justification, study design, analysis populations, confidence intervals, missing data, adherence, applicability, resource quantities, unit costs, total and incremental costs, clinical and cost uncertainty, shifted or excluded costs, limitations and the conditional decision conclusion.

Sources

• Briggs AH, O'Brien BJ. The death of cost-minimization analysis? Health Economics. 2001;10(2):179-184.
• Dakin H, Wordsworth S. Cost-minimisation analysis versus cost-effectiveness analysis, revisited. Health Economics. 2013;22(1):22-34.
• Wailoo A, Dixon S. The use of cost minimisation analysis for the appraisal of health technologies. NICE Decision Support Unit report. 2019.
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). Published 2022; updated 2026.
• Pharmaceutical Benefits Advisory Committee. Guidelines for preparing a submission to the Pharmaceutical Benefits Advisory Committee, version 5.0. Australian Government Department of Health. 2016. Section 3B (Cost minimisation).
• Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes, 4th edition. Oxford University Press. 2015.