ICER: The Incremental Cost-Effectiveness Ratio and Thresholds

Almost every reimbursement debate about a new medicine eventually reaches the same number: the incremental cost-effectiveness ratio, or ICER. It compresses an entire economic model into one figure, the extra cost of each extra unit of health, most often written as cost per QALY gained.

It is the headline result of most cost-utility analyses submitted to HTA bodies, yet on its own it says little. Whether £30,000 per QALY is good value depends on the comparator, the quadrant of the cost-effectiveness plane, the threshold and the uncertainty. This article works through each.

The ICER formula, step by step

Label the new intervention 1 and its comparator 0. With C for cost and E for health effect, the incremental cost-effectiveness ratio is (C₁ − C₀) divided by (E₁ − E₀), often shortened to ΔC / ΔE.

Costs are usually total discounted costs per patient over the model’s time horizon. Effects are most often quality-adjusted life years, explained in our guide to what a QALY is and how it is calculated, but life years or a clinical unit such as cases avoided can also be used. Both figures are usually averages from a model, which is why economic modelling with decision trees and Markov models shapes the ICER so strongly. Our guide to Markov models in health economics shows how a model turns transition probabilities into costs and QALYs.

A worked example with hypothetical numbers

Take a hypothetical new drug for a chronic condition, compared with current standard care. The figures are round numbers chosen for illustration and do not describe any real product.

The ICER is £30,000 divided by 1.0 QALY, or £30,000 per QALY gained.

Two points are easy to miss. The ratio is built from differences, so a drug with very high total costs can still have a moderate ICER if the comparator is also expensive. And the denominator is powerful: halve the QALY gain to 0.5 and the same extra cost gives £60,000 per QALY. Small, uncertain health gains make ICERs volatile. To test your own numbers, use our free ICER calculator, which also shows the quadrant of the cost-effectiveness plane and the net monetary benefit.

Reading the cost-effectiveness plane

The cost-effectiveness plane plots incremental health effect on the horizontal axis and incremental cost on the vertical axis, with the comparator at the origin. Where the new option lands decides how, or whether, its ICER should be read.

Why negative ICERs are not reported

Suppose one option saves £5,000 and gains 0.5 QALYs, while another costs £5,000 more and loses 0.5 QALYs. Both produce an ICER of −£10,000 per QALY, yet the first is clearly better than its comparator and the second clearly worse. Because the sign hides which quadrant you are in, the usual recommendation is not to report negative ICERs at all and to use the labels “dominant” and “dominated” instead.

The south-west quadrant needs similar care. There, an ICER of £50,000 per QALY means £50,000 saved for every QALY lost, so a result above the threshold, not below it, favours the cheaper option. The arithmetic is identical; the interpretation is inverted, and the result should be labelled explicitly.

Dominance and extended dominance with several options

With more than two options, analysts run a full incremental analysis: rank the options by cost, remove any that are dominated, then compare each remaining option with the next cheaper one still in contention. The survivors form the cost-effectiveness frontier.

Simple dominance. D costs more than C and produces fewer QALYs, so it is dominated and drops out first.

Extended dominance. In cost order, B versus A gives £10,000 / 0.2 = £50,000 per QALY, and C versus B gives £10,000 / 0.8 = £12,500 per QALY. The ratio falls as you move to the more expensive option, which is the signature of extended dominance: B has a higher ICER than a more effective alternative. Remove B and compare C directly with A: £20,000 / 1.0 = £20,000 per QALY.

Why not report B’s ratio? Anyone willing to pay £50,000 per QALY to move from A to B would buy each extra QALY more cheaply by moving to C. Treating half of patients with A and half with C would cost £20,000 per patient on average and yield 5.5 QALYs, more than B for the same money. Ratios for dominated options invite comparisons no sound decision should rest on, which is also why calculating every option’s ICER against standard care alone is a common error: it hides extended dominance.

What is a cost-effectiveness threshold?

An ICER becomes a decision only when it is set against a cost-effectiveness threshold, often written as the Greek letter λ (lambda) and also called the willingness-to-pay threshold. For a result in the north-east quadrant, an ICER below the threshold suggests the technology is good value; an ICER above it suggests the health gained costs more than the system is prepared to pay.

λ can be read in two ways: what a society is willing to pay for a unit of health, or how much health the same money would produce in existing services, which ties it to the opportunity cost reasoning at the heart of HTA. The two can give very different numbers, and every published threshold is partly a policy choice.

Thresholds are rarely applied as hard cut-offs. Committees weigh the ICER alongside the certainty of the evidence and any benefits the QALY may not capture, within the wider appraisal process described in our introduction to what health technology assessment is.

ICER thresholds by country: NICE, the US and WHO

Health systems differ in whether they publish an explicit cost per QALY reference value at all. The table summarises the position as of October 2026 for the systems verified for this article.

England: the NICE threshold from April 2026

On 1 December 2025 the UK government announced, and NICE confirmed, that the standard cost-effectiveness threshold for technology appraisals would rise from £20,000–£30,000 to £25,000–£35,000 per QALY gained from April 2026, covering new appraisals and those already under way. NICE expects the change to add around 3 to 5 recommended medicines or indications a year. The Office of Health Economics notes that the rise formed part of a wider UK–US agreement on medicines announced the same month.

NICE states that its manual, last updated on 31 March 2026, now reflects the new range. So far the change applies only to technology appraisals; NICE knows of no plans to alter the separate range for highly specialised technologies (£100,000–£300,000 per QALY gained). In March 2026 the Department of Health and Social Care also confirmed that ministers would gain a narrow power to direct the threshold NICE uses.

The range interacts with NICE’s severity modifier, introduced in 2022 to replace the end-of-life criteria. QALYs gained in severe conditions can be weighted by 1.2 or 1.7, depending on the proportional or absolute QALY shortfall patients face, which works like a higher threshold. NICE plans to review how its move to a new EQ-5D-5L value set affects which appraisals qualify.

United States: ICER’s health-benefit price benchmarks

The Institute for Clinical and Economic Review does not set coverage rules. Under its 2023 Value Assessment Framework, still listed as current on its website in October 2026, it calculates a health-benefit price benchmark: the net price, after rebates and discounts, at which a treatment would reach $100,000 to $150,000 per QALY and per equal value life year (evLY) gained. Its reports also show results at $50,000, $100,000, $150,000 and $200,000. How its reports fit alongside Medicare and private payers is covered in our guide to health technology assessment in the US.

The evLY (evLYG in earlier frameworks) counts any extension of life equally, whatever the patient’s health during those years, answering concerns that QALYs undervalue life extension for people with chronic illness or disability.

WHO-CHOICE and the GDP per capita rule of thumb

In the early 2000s the WHO Commission on Macroeconomics and Health and the World Health Report 2002 put forward benchmarks later promoted through WHO-CHOICE: an intervention costing less than one times GDP per capita per DALY averted was labelled highly cost-effective, and less than three times cost-effective.

The heuristic has been heavily criticised. Marseille and colleagues argued in 2015 that it is so easily met that it barely discriminates between options, rests on untested assumptions about what countries are willing to pay, and ignores affordability entirely. In 2016, WHO staff writing in the Bulletin of the World Health Organization concluded that GDP-based thresholds lack country specificity and, combined with model uncertainty, can lead to poor spending decisions. They recommended weighing cost-effectiveness alongside budget impact and feasibility in a transparent, country-led process. WHO-CHOICE likewise calls cost-effectiveness an important criterion, but not the only one.

Germany and systems without an explicit threshold

Germany’s AMNOG process does not use a cost per QALY threshold for market access. IQWiG describes economic evaluation as playing no significant role in the early benefit assessment: reimbursement depends on added benefit over an appropriate comparator therapy, and without demonstrated added benefit statutory insurers pay no more than for the comparator. IQWiG’s methods have also long rejected a single threshold across all indications, using an efficiency frontier within each therapeutic area when a formal cost–benefit assessment is done. For other countries, check the agency’s current methods guide rather than figures quoted second-hand.

Net monetary benefit: a cleaner decision rule

Net monetary benefit (NMB) converts the health gain into money using the threshold, then subtracts the extra cost. A positive incremental NMB means the new option is cost-effective at that threshold.

Return to the hypothetical drug, with ΔE of 1.0 QALY and ΔC of £30,000. At λ = £25,000 the NMB is −£5,000; at λ = £35,000 it is +£5,000. Same model, opposite verdicts: where a result falls within a threshold range matters.

Why many analysts prefer net benefit

• The sign is unambiguous in every quadrant, so there is no negative-ICER problem.
• With many options, the one with the highest expected NMB is the most efficient choice, without a step-by-step frontier analysis.
• NMB is linear, so it can be averaged across thousands of simulations, which ratios cannot be reliably.

Some methodologists, such as Paulden in a 2020 PharmacoEconomics paper, go further and argue that reporting ICERs is unnecessary once net benefit is available. HTA bodies such as NICE still frame their thresholds around ICERs, so analysts need to be fluent in both.

Uncertainty: PSA and acceptability curves

A model’s ICER is a point estimate. Probabilistic sensitivity analysis (PSA) draws uncertain inputs from probability distributions and reruns the model many times, and NICE prefers the base-case result to come from a probabilistic analysis where possible. The cost-effectiveness acceptability curve (CEAC) then plots the probability that the option is cost-effective across a range of thresholds. Tornado diagrams, scenario analyses and value of information are covered in the module on handling uncertainty with PSA and CEACs.

How to read an ICER critically

An incremental cost-effectiveness ratio is only as good as the decision problem and model behind it. These are the questions experienced reviewers ask first.

The course practises spotting these issues in the practitioner’s workshop on reading submissions and red flags.

Cost-effective is not the same as affordable

An ICER is a per-patient efficiency measure. It says nothing about how many patients will be treated or what the total bill will be, so a technology with an acceptable ICER for a large population can still strain a fixed budget.

That affordability question is answered by a separate analysis, explained in our guide to budget impact analysis and how it differs from cost-effectiveness. HTA bodies usually look at both. If you are mapping which roles produce these analyses and which review them, see HEOR vs HTA vs market access.