Cost-Effectiveness Analysis (CEA)
- Introduced by
- Drummond & Stoddart (Health Economics Research Group, McMaster University)
- Year
- 1984
- Type
- Method
- Family
- economic evaluation framework
What does it do?
Cost-effectiveness analysis compares the incremental cost per unit of health benefit gained by one intervention relative to a comparator (standard care or best alternative).
Developed rigorously in the 1980s by Drummond, Stoddart, and colleagues, CEA is now the standard framework for technology appraisal globally. NICE, HAS, CADTH, and other health technology assessment bodies use CEA to decide which treatments warrant public funding and at what price.
The intuition behind the method
A healthcare system has finite resources. Drug A costs $100,000 and extends survival 1 year; Drug B costs $50,000 and extends survival 1 year: clearly Drug B is better (same benefit, lower cost). But what if Drug A extends survival 3 years and costs $100,000, while Drug B extends survival 1 year and costs $50,000? Drug A costs $33,333 per year; Drug B costs $50,000 per year—now A looks better. CEA formalizes this trade-off: the Incremental Cost-Effectiveness Ratio (ICER) = extra cost per extra unit of benefit. If a society is willing to pay $50,000 per life-year (or QALY), then interventions with ICER < $50,000 are funded; those with ICER > $50,000 are not. This enables systematic, transparent priority-setting.
History
Formal CEA methodology crystallized in the 1980s. Drummond & Stoddart (1984) published the first textbook on economic evaluation of health programs. Gold et al. (1996) published 'Cost-Effectiveness in Health and Medicine,' establishing U.S. reference guidelines. Drummond et al. (2005, 2015 4th edition) expanded internationally accepted methods. NICE (UK) adopted CEA-based appraisal in 1999, with QALY as primary metric. By the 2000s, CEA was standard in Australia (PBS), Canada (CADTH), France (HAS), Germany (IQWiG), and Nordic countries. WHO endorsed cost-effectiveness thresholds in 2001 (1–3× GDP per capita per DALY, later updated). Today, CEA is integrated into HTA everywhere; debate centers on thresholds, equity adjustments, and inclusion of societal costs (productivity) vs healthcare-only costs.
When to use it
When comparing two or more health interventions (drugs, devices, procedures, programs) and decision-makers need evidence on value for money. CEA is mandatory for health technology appraisal submissions in most developed countries (NICE, HAS, CADTH, PMPRB in Canada, PBS in Australia). Use when: a new treatment has proven efficacy; cost is not negligible; comparison is against current standard care; effectiveness can be measured in generic units (QALYs, life-years, DALYs); a transparent, replicable decision framework is needed. Not appropriate for: comparing non-health interventions (education, employment); situations where cost is not a constraint (rare); when effectiveness is uncertain and unmeasurable.
Limitations and considerations
- Threshold choice is arbitrary: who decides what society is willing to pay per QALY? £20,000 vs £30,000 vs £50,000? Different countries, stakeholders, and experts disagree; threshold shifts with political will.
- Assumes comparability of QALYs: a QALY to a young, healthy person is treated the same as a QALY to an elderly disabled person (equity concern).
- Data-dependent: ICER hinges on trial-derived efficacy, cost estimates, and real-world follow-up. If RCT is small, biased, or from different population, ICER is uncertain.
- Time horizon matters significantly: a 10-year CEA produces different conclusion than lifetime CEA. Choice is normative (policy decision), not empirical.
- Discounting: future costs and benefits discounted at 3–5%, which favors interventions with immediate impact; controversial for long-term prevention (e.g., climate health) or childhood interventions.
Source conditions do not replace independent scientific review or validation by an analysis engine.
Strengths
- Enables transparent, systematic comparison of interventions across disease areas using a common metric (cost per unit of benefit), supporting rational resource allocation.
- Incorporates both costs and benefits, preventing overinvestment in expensive low-benefit treatments or underinvestment in cheap, highly effective treatments.
- Flexible: effectiveness measured in QALYs, life-years, DALYs, or disease-specific outcomes; can accommodate multiple perspectives (healthcare system, societal).
- Facilitates negotiation: pharmaceutical companies and payers can use CEA to justify pricing; transparent ICER enables evidence-based price discussion.
Applications
- Pharmaceutical pricing negotiation: a pharmaceutical company demonstrates ICER < £20,000/QALY to NICE; NICE recommends; price is set to achieve target ICER.
- Device appraisal: surgical device (e.g., hip prosthesis) CEA identifies which materials/designs are most cost-effective, informing procurement guidelines.
- Public health program evaluation: vaccination program CEA estimates cost per DALY averted; compared to threshold guides investment scale.
- Clinical practice guideline setting: multiple drugs for hypertension compared via CEA; guideline recommends most cost-effective first-line, conditional recommendations for alternatives.
- Research funding priorities: Disease X has high disease burden but no cost-effective treatment; funding agencies prioritize research into cheaper treatments or prevention.
How it works
How it works
Step 1: Define the health question—e.g., 'Is Drug X cost-effective vs standard chemotherapy for lung cancer?' Step 2: Choose the comparator (almost always best current care, not placebo). Step 3: Identify the patient population and time horizon (e.g., 10-year lifetime perspective for cancer). Step 4: Determine the effectiveness measure: QALYs (most common), life-years, disease-specific outcomes (symptoms resolved, hospitalizations prevented). Step 5: Identify costs by category: (a) direct medical costs (drug acquisition, administration, monitoring, hospitalizations, adverse event management); (b) direct non-medical costs (patient transport, caregiver time); (c) productivity costs (lost work days, early retirement). In practice, many analyses focus on direct medical costs; inclusion of productivity costs is debated. Step 6: Data sources: RCT for efficacy, published epidemiology for disease progression, hospital costing studies for resource use, price lists for unit costs. Step 7: Build a decision model (see markov-model-health-economics) to project costs and outcomes over time.
Formula
Define comparator (standard care or placebo), time horizon (years of follow-up), and effectiveness measure (QALYs, DALYs, clinical outcomes). Identify and quantify costs: direct medical (drug, hospitalization, staff), direct non-medical (transport, care), indirect (lost productivity). Obtain efficacy data from RCTs or observational studies. Build decision model (Markov, cohort simulation, or decision tree).
Formula: ICER = (Cost_new − Cost_standard) / (Effect_new − Effect_standard). Example: A new cancer drug costs $120,000; standard chemotherapy costs $30,000. Drug extends median survival 2 years; standard chemotherapy extends survival 1.5 years. Cost difference = $90,000; Benefit difference = 0.5 years. ICER = $90,000 / 0.5 = $180,000 per life-year. If benefit is measured in QALYs (utility-weighted): Drug gains 1.8 QALYs; standard gains 1.2 QALYs. ICER = $90,000 / 0.6 QALYs = $150,000 per QALY. Quadrant analysis: (1) NE (northeast): new treatment costs more, gains more health. ICER > 0. Compare to threshold: if $150,000/QALY < societal WTP threshold, adopt. (2) SE (southeast): new treatment costs less, gains more health. ICER < 0. Dominant—always adopt. (3) NW (northwest): new treatment costs less, gains less health. ICER < 0. Rarely adopted (worse outcomes for savings). (4) SW (southwest): new treatment costs more, gains less health. ICER > 0. Reject—strictly dominated. Net Benefit calculation (alternative to ICER): NB = (Effect × WTP) − Cost. Positive NB = cost-effective at that WTP threshold.
Formula
Calculate total cost and total effectiveness for intervention and comparator. ICER = (Cost_intervention − Cost_comparator) / (Effect_intervention − Effect_comparator). Interpret quadrant: NE (more costly, more effective) vs comparator, apply threshold; SE (less costly, more effective) = dominant; NW (less costly, less effective) = rarely chosen; SW (more costly, less effective) = reject.
Acceptability depends on threshold. NICE (UK): £20,000–£30,000 per QALY considered cost-effective (based on estimated value of care per QALY in NHS). WHO (World Health Organization, developing countries): 1–3 × GDP per capita per DALY averted is cost-effective (threshold varies by country wealth). CADTH (Canada): typically uses $30,000–$50,000 per QALY CAD. If ICER < threshold, intervention is cost-effective and recommended for funding/reimbursement. If ICER > threshold, typically not recommended unless special circumstances (rare disease, significant uncertainty, unmet need). Sensitivity analysis is critical: vary key assumptions (drug price, efficacy, discount rate, time horizon) and recalculate ICER. Probabilistic sensitivity analysis (Monte Carlo simulation) produces a cloud of plausible ICERs and an acceptability curve showing P(cost-effective) at each WTP threshold. Subgroup analysis: does ICER differ for elderly vs young, severe vs mild disease? Heterogeneity may justify restricted recommendations (e.g., fund Drug X only for elderly patients).
Formula
Compare ICER to willingness-to-pay threshold (WTP). If ICER < WTP, intervention is cost-effective and recommended. If ICER > WTP, intervention is not cost-effective. Conduct deterministic and probabilistic sensitivity analysis to test robustness. Report acceptability curve: probability intervention is cost-effective across range of WTP values. Subgroup analysis by age, disease severity, or risk profile to identify who benefits most.
Common pitfalls
- Ignoring the comparator: comparing new drug vs placebo instead of current standard treatment. This overstates benefit (CEA should always ask, 'vs what?'). FDA approval vs placebo is not the same as CEA vs standard care.
- Using inappropriate effectiveness measures: CEA should use generic measures (QALY, life-years) that enable cross-disease comparison, not disease-specific surrogate endpoints (e.g., CD4 count for HIV). Some analyses mix clinical outcomes (life-years) with utility measures (QALYs) improperly.
- Forgetting costs: focusing only on efficacy (clinical outcomes) without quantifying true incremental cost, producing misleading conclusions about value.
- Cherry-picking thresholds: presenting ICER as cost-effective using a high WTP (£50,000/QALY) when NICE threshold is £20,000. Sensitivity analysis should test range of thresholds transparently.
- Failing to account for parameter uncertainty: reporting a single ICER (point estimate) without confidence intervals or sensitivity ranges, giving false precision.
- Ignoring equity: CEA is utilitarian (most health per dollar); it does not account for equity (Should sicker patients get priority? Should costly drugs be subsidized for low-income patients?). Policy should incorporate equity alongside efficiency.
Frequently asked questions
What is the difference between CEA and CBA (cost-benefit analysis)?
CEA measures benefit in natural units (QALYs, life-years); all benefits are in the same units, summed. CBA assigns monetary value to benefits, so both cost and benefit are in dollars (or pounds). CEA: Cost / Health Benefit (e.g., $50,000 per QALY). CBA: Net Benefit = Monetary Benefit − Cost (e.g., $2 million − $1 million = $1 million net benefit). CEA is more common in healthcare because assigning a dollar value to health is ethically fraught ('How much is a life worth?'). CBA is standard in transportation, environmental policy.
Why do different countries use different cost-effectiveness thresholds?
Thresholds reflect a country's healthcare resources and priorities, not a universal truth. UK (£20,000–£30,000/QALY) has higher threshold than lower-income countries (WHO: 1–3× GDP per capita) because higher GDP allows more spending per health benefit. A QALY in a $50,000-GDP country (e.g., Vietnam) is worth less by purchasing power. Threshold also reflects political will: if a country increases health spending, threshold can rise.
How are costs adjusted for inflation and currency differences across studies?
CEA studies should report costs in a specified year and currency. If comparing studies from different years, inflate/deflate using health price index (more accurate than general CPI, as healthcare costs diverge). For international comparison, use Purchasing Power Parity (PPP) exchange rates, which adjust for cost-of-living differences, not nominal exchange rates.
What discount rate should be used for future costs and benefits?
Standard practice: 3–5% annually in developed countries (UK, Canada, Australia, USA typically use 3%; some use 5%). The logic: society values money and health today more than in 10 years. A benefit in year 10 is discounted to ~60% of present value at 3%. Sensitivity analysis should test 0%, 3%, 5% to see if results are robust. Debate: some argue lower discount rates (0–1%) for long-term prevention (climate, child health) to avoid excessive discounting of future generations.
How is CEA conducted when there are multiple outcomes (mortality, symptom relief, adverse effects)?
Consolidate into single metric: QALY (utility-weighted; each health state assigned utility 0–1, reflecting symptom relief, side effects, mortality risk). Build a decision model tracking transitions between health states, incorporating improvements (symptom relief) and harms (side effects). If multiple outcomes cannot be combined into QALY, report separately (multi-outcome CEA); decision-makers integrate trade-offs.
What is the 'perspective' of a CEA, and why does it matter?
Perspective defines whose costs count. Healthcare perspective: only medical costs (drugs, hospital, staff) count. Societal perspective: includes productivity costs (lost wages, caregiver time) and non-medical costs. Healthcare CEA is narrower; societal CEA is broader and often shows larger benefit if intervention also enables return-to-work. Different countries prefer different perspectives: NICE uses healthcare perspective; Australian PBS includes societal; WHO recommends societal for developing countries. Analysts should pre-specify and clearly report perspective, as it affects ICER.
Related methods
Sources
Gold, M. R., Siegel, J. E., Russell, L. B., & Weinstein, M. C. (Eds.). (1996). Cost-Effectiveness in Health and Medicine. New York: Oxford University Press.
Open publication referenceCopied.Drummond, M. F., Sculpher, M. J., Claxton, K., Stoddart, G. L., & Torrance, G. W. (2015). Methods for the Economic Evaluation of Health Care Programmes (4th ed.). Oxford: Oxford University Press.
Open publication referenceCopied.National Institute for Health and Care Excellence (NICE). (2022). Guide to the Methods of Technology Appraisal. London: NICE.
Open publication referenceCopied.