Cost-effectiveness thresholds (CETs) are fundamental to health technology assessment and resource allocation. For decades, many countries have relied – explicitly or implicitly – on the heuristic rule of one to three times gross domestic product per capita (GDPpc), initially proposed by the World Health Organization (WHO) (1), to determine whether interventions represent good value for money (originally conceived as a broad investment benchmark rather than a formal decision threshold, the WHO rule has nevertheless continued to influence economic evaluations, reimbursement discussions, and published literature in many settings, particularly in low- and middle-income countries). Although CETs should not be the sole criterion for decision-making, they should be used to inform resource allocation and HTA deliberations alongside considerations such as equity, severity, affordability, and feasibility (Reference Espinosa, Rodríguez-Lesmes and Romano2). In public healthcare systems with constrained budgets and a goal of improving population health, CETs should reflect the opportunity costs of health spending – that is, the health that is forgone elsewhere from displacing resources to fund a new intervention.

Health gains associated with the new intervention can then be compared to the health loss expected from disinvestment to estimate the net health benefits from the new intervention. Growing empirical evidence based on real-world data challenges the validity of the 1–3×GDPpc rule as a CET, suggesting that the true opportunity cost of health spending is often considerably lower (Reference Ochalek, Lomas and Claxton3). Funding new interventions based on 1–3×GDPpc may, therefore, result in significant net population health losses. Lower empirical CET estimates imply more conservative funding decisions, as fewer interventions would be considered cost-effective compared with decisions based on higher thresholds such as the heuristic rule. This reflects the possibility that healthcare resources displaced elsewhere in the system may generate larger population health gains than previously assumed.

In this letter, we compare international empirical estimates of CETs representing health opportunity costs with CETs based on the 1–3×GDPpc rule. We conducted a literature search in PubMed, Scopus, Embase, EconLit, and Google Scholar using the following keywords: “empirical cost-effectiveness threshold,” “marginal productivity health expenditure,” “elasticity health expenditure,” “health opportunity cost,” and “QALY/DALY/YLL elasticity expenditure real data.” We identified and reviewed 10 econometric studies estimating the costs of health spending from the supply-side perspective, that is, based on the marginal productivity of health expenditure (estimates for Saudi Arabia (Reference Al-Jedai, Lomas and Almudaiheem4), the United States (Reference Vanness, Lomas and Ahn5), and Germany (Reference Gandjour6) were excluded as they did not use real-world data, did not directly estimate the marginal cost per unit of health produced, or did not apply an econometric analysis to infer a causal effect. Maximum CETs values of 0.86, 1.62, and 2.18 GDPpc per QALY were reported, respectively). Citations, cross-references, and the AI tool Connected Papers were also used to identify additional relevant studies.

For each study, we extracted the country, income-level classification according to World Bank guidelines, econometric method applied, type of included health expenditure, main point value of the CET, and its relation as a percentage of GDPpc. If this information was not explicitly reported in the study, it was estimated using the per capita GDP in local currency for the base year in which the CET results are presented obtained from national statistical institutes or ministries of finance. In addition, we estimated the net health impact associated with the approval of new health technologies under CET of 1× or -3×GDPpc, using the methodology proposed by Murphy et al. (Reference Murphy, Griffin and Walker7) (Health (QALYs) lost per million USD was estimated by comparing the health expected from funding technologies at CETs equivalent to 1× or 3×GDPpc with the health that could be generated if the same resources were allocated according to empirically estimated health opportunity costs. Calculations assume that approved technologies have an ICER equal to the respective GDP-based threshold and that displaced healthcare spending generates health at the empirically estimated CET (Reference Murphy, Griffin and Walker7)). No studies from low-income or lower-middle-income countries met the predefined inclusion criteria. Therefore, all included countries were classified as either high-income or upper-middle-income countries.

Across all countries, none of the empirically estimated CETs reached the upper bound of the WHO heuristic (3×GDPpc). Indeed, in all but two countries, the estimates were below GDPpc, especially among upper-middle-income countries (see Table 1). Considerable heterogeneity was observed across countries and methods, with instrumental variables emerging as the most commonly applied econometric technique for estimating empirical CETs, reflecting its suitability for addressing endogeneity in health expenditure analyses (these estimates should not be interpreted as fixed national decision thresholds but as empirical approximations of marginal productivity under specific methodological and institutional assumptions). Most studies examine the marginal productivity of public healthcare spending, although one adopted a broader perspective by considering total health expenditure including out-of-pocket payments (China), another took a narrower approach, focusing solely on hospital expenditure (Netherlands), and one specifically analyzed the impact of statutory health insurance spending (Colombia).

Table 1.

Comparative synthesis of empirical studies on cost-effectiveness thresholds by country

Table 1. Long description

The table contains 10 columns: Country, Income level 2025, Econometric method, Health expenditure data, C E T, G D P p c, Year base, C E T all over G D P p c approx, Health Q A L Y s lost per million U S D with C E T equals 1 times G D P p c (C E T equals 3 times G D P p c), and Reference.

Data rows:

* Indonesia: Upper-middle income; Dynamic panel; Public spending; C E T U S D 235 per D A L Y averted; G D P p c U S D 4,107; Year 2019; Ratio 0.06; Health lost -4,000 (-4,170).

* Australia: High income; Instrumental variables; Public spending; C E T A U D 28.033 per Q A L Y gained; G D P p c A U D 63,499; Year 2011; Ratio 0.44; Health lost -19 (-29).

* England: High income; Instrumental variables; Public spending; C E T 12.936 pounds per Q A L Y gained; G D P p c 25,916 pounds; Year 2008; Ratio 0.50; Health lost -21 (-35).

* South Africa: Upper-middle income; Fixed-effects panel; Public spending; C E T Z A R 38.500 per D A L Y averted; G D P p c Z A R 72,700; Year 2015; Ratio 0.53; Health lost -156 (-273).

* China: Upper-middle income; Ordinary least squares; Total spending; C E T R M B 37.446 per D A L Y averted; G D P p c R M B 59,438; Year 2017; Ratio 0.63; Health lost -67 (-143).

* Sweden: High income; Instrumental variables; Public spending; C E T S E K 367.507 per L Y gained; G D P p c S E K 444,000; Year 2016; Ratio 0.83; Health lost -4 (-17).

* Colombia: Upper-middle income; Instrumental variables; Statutory insurance; C E T C O P 17 million per Q A L Y gained; G D P p c C O P 19.7 million; Year 2019; Ratio 0.86; Health lost -27 (-138).

* Spain: High income; Fixed-effects panel; Public spending; C E T 27.165 Euro per Q A L Y gained; G D P p c 28,162 Euro; Year 2022; Ratio 0.96; Health lost -2 (-24).

* Greece: High income; Instrumental variables; Public spending; C E T 27.117 Euro per Q A L Y gained; G D P p c 17,270 Euro; Year 2019; Ratio 1.57; Health lost 19 (-16).

* The Netherlands: High income; Fixed-effects panel; Hospital spending; C E T 73.600 Euro per Q A L Y gained; G D P p c 40,240 Euro; Year 2014; Ratio 1.83; Health lost 8 (-4).

Comparison between national empirical estimates and 1–3×GDPpc CETs reveals a misalignment between national realities and global heuristics. Although the WHO’s 1–3×GDPpc CET was originally conceived as a broad investment benchmark rather than a decision threshold, its widespread misapplication has often led to inefficient allocation of scarce health budgets. The WHO has since clarified that such ratios should not be interpreted as normative decision rules (Reference Bertram, Lauer and De Joncheere18).

Respecting the comparative analysis of health loss per million dollars invested, we observe significant variation across countries, closely linked to income levels and the relative value of empirical CETs compared to GDPpc multipliers. In all countries, the application of a 3×GDPpc threshold results in net health losses, highlighting the limitations of applying this to inform decisions, especially in middle-income contexts where marginal returns on healthcare spending are higher.

Evidence from England illustrates the potential harm of applying inflated thresholds. Naci et al. (Reference Naci, Murphy, Woods, Lomas, Wei and Papanicolas19) estimated that between 2000 and 2020, drugs approved by the National Institute for Health and Care Excellence under their decision threshold (£ 20.000 – £ 30.000 per QALY) generated approximately 3·75 million quality-adjusted life years (QALYs) but displaced existing services equivalent to 5 million QALYs based on empirically estimated health opportunity costs of £ 15.000 per QALY (Reference Claxton, Martin and Soares10; Reference Lomas, Martin and Claxton20) – a net population health loss of 1·25 million QALYs. Funding decisions made against 1–3×GDPpc (£ 25.916 – £ 77.748) would have resulted in greater loss of population health.

Empirically derived CETs grounded in opportunity cost provide a more coherent and ethically defensible framework for priority setting. These thresholds quantify how much health is forgone when new interventions are adopted and are, therefore, essential for health systems that aim to ensure that the introduction of new technologies improves overall population health. However, their estimation requires robust administrative and patient-level data, which remain limited in many low- and middle-income settings. Strengthening national data systems and analytical capacity is, therefore, critical for estimating evidence-based thresholds.

A further methodological issue concerns the periodic adjustment of thresholds. Vallejo-Torres (Reference Vallejo-Torres15) demonstrated that updating CETs by inflation alone yields implausible results, as the marginal productivity of health spending evolves differently from consumer prices. We recommend that updated estimates of health opportunity costs should be based on current data estimating the marginal productivity of health spending over time, incorporating novel statistical methods as they become available. Where this is not feasible, updated estimates can be generated using current data on average healthcare spending and health outcomes (see Equation 2 in (Reference Espinosa, Rodríguez-Lesmes and Orozco14)) assuming a constant elasticity of health expenditure over time.

We contend that global health decision-making must now move beyond the mechanical application of the 1–3×GDPpc rule. Countries should adopt context-specific, empirically grounded thresholds that reflect local fiscal realities and opportunity costs. Such thresholds should inform deliberative processes that also integrate ethical and distributional considerations. Even where these are not yet formally adopted by countries, researchers should apply CETs representing health opportunity costs in published cost-effectiveness analyses. Journal editors, peer-reviewers, and research funding agencies have a role to play in actively encouraging this.

Transitioning from global heuristics to empirical evidence is an important step toward improving the efficiency and accountability of health resource allocation.