Cost per QALY

Cost per QALY in Incremental Analysis and League Tables

Results of a cost-utility analysis are usually summarised as a cost-per-QALY figure, but the same label is attached to calculations that answer different questions. A cost per QALY only supports a funding decision when it prices the extra quality-adjusted life years bought by moving from the next best option, under stated conventions. That ratio is the QALY form of the incremental cost-effectiveness ratio, and an illustrative example shows how an average ratio, or a comparison with the wrong option, can make a poor-value treatment look like good value. League tables, NICE's range of £25,000 to £35,000 per QALY gained under its manual (PMG36), the reference-case conventions behind a reported figure and negative ratios all bear on how a reported figure should be read.

Cost per QALY gained as the QALY form of the ICER

In a cost-utility analysis health effects are measured in QALYs, so the incremental ratio is expressed in money per QALY gained. NICE's manual (PMG36) states that the ratios reported must be the ratio of expected additional total cost to expected additional QALYs compared with alternative technologies. For a technology T compared with a comparator S, the ratio is as follows.

$$\text{Cost per QALY gained}_{T,S} = \frac{\bar{C}_T - \bar{C}_S}{\bar{Q}_T - \bar{Q}_S} = \frac{\Delta C}{\Delta Q}$$

where $\bar{C}_T$ and $\bar{C}_S$ are the expected total costs of T and S, $\bar{Q}_T$ and $\bar{Q}_S$ are their expected total QALYs, $\Delta C$ is the incremental cost and $\Delta Q$ is the incremental QALY gain, all discounted and measured over the same time horizon.

The ratio belongs to a pair of options, so a figure quoted without its comparator cannot be interpreted. It is also a ratio of expected values, which generally differs from an average of ratios taken from separate simulation runs. The PMG36 manual describes the QALY as the most appropriate generic measure of health benefit reflecting both mortality and quality-of-life effects, which is what lets figures from different diseases be set side by side. Cost per QALY gained and incremental cost-utility ratio name the same quantity; the ICER page covers the general calculation and the cost-effectiveness plane.

Average cost per QALY answers a different question

An average cost per QALY divides the total cost of one option by the total QALYs experienced under it. It describes that option on its own, without reference to what the same patients would have cost and experienced under an alternative. Written formally, it is as follows.

$$\text{ACER}_T = \frac{\bar{C}_T}{\bar{Q}_T}$$

where $\text{ACER}_T$ is the average cost-effectiveness ratio of option T, and $\bar{C}_T$ and $\bar{Q}_T$ are its expected total cost and expected total QALYs.

The average ratio prices every QALY that patients live under T, including those they would have lived anyway under the alternative, while the incremental ratio prices only the QALYs that the decision adds. A health system choosing between options pays for the difference between them, so only the incremental ratio, or an equivalent net benefit calculation, shows whether the extra spending buys enough extra health. The two ratios are equal only when the comparator has the same average ratio as T, that is when $\bar{C}_S / \bar{Q}_S = \bar{C}_T / \bar{Q}_T$. A comparator with zero cost and zero QALYs is a special case of this, but it almost never describes real patients, since untreated people still live and still use care.

Worked example with three cost-per-QALY figures for one treatment

The figures below are illustrative. Three mutually exclusive options for a chronic condition are compared over a lifetime horizon, with costs and QALYs already discounted.

1. Average ratios. Usual care costs £4,000 / 6.00 = £666.67 per QALY, treatment A costs £10,000 / 6.50 = £1,538.46 per QALY and treatment B costs £22,000 / 6.80 = £3,235.29 per QALY. Read on their own, all three look far below any threshold.

2. Treatment B against usual care only. Leaving out treatment A gives the following.

$$\frac{22000 - 4000}{6.80 - 6.00} = \frac{18000}{0.80} = 22500$$

where the result is the cost in pounds per QALY gained by moving from usual care to B, which appears to fall below £25,000.

3. Fully incremental analysis. Options are ranked by cost and each is compared with the next less costly option still in the analysis. Moving from usual care to A gives the following.

$$\frac{10000 - 4000}{6.50 - 6.00} = \frac{6000}{0.50} = 12000$$

where the result is the cost in pounds per QALY gained by moving from usual care to A. Moving from A to B gives the following.

$$\frac{22000 - 10000}{6.80 - 6.50} = \frac{12000}{0.30} = 40000$$

where the result is the cost in pounds per QALY gained by moving from A to B. The ratios rise with cost, so no option is ruled out by extended dominance, and the step from A to B costs £40,000 per QALY gained, above £35,000.

4. Confirmation with net health benefit. Net health benefit converts costs into QALYs at a stated value per QALY, as follows.

$$NHB_i = \bar{Q}_i - \frac{\bar{C}_i}{\lambda}$$

where $NHB_i$ is the net health benefit of option $i$ in QALYs, $\bar{Q}_i$ and $\bar{C}_i$ are its expected QALYs and costs, and $\lambda$ is the value placed on one QALY. At £25,000 per QALY the net health benefits are 5.84 for usual care (6.00 − 4,000 / 25,000), 6.10 for A (6.50 − 10,000 / 25,000) and 5.92 for B (6.80 − 22,000 / 25,000). At £35,000 per QALY they are about 5.886, 6.214 and 6.171, so A has the highest net health benefit at both values.

Treatment B has now been described as costing £3,235, £22,500 or £40,000 per QALY. Only the last figure answers the decision question, because a health system that can choose A must judge B by what it adds over A, which is £12,000 more per patient for 0.30 extra QALYs, which at £25,000 per QALY would displace 0.48 QALYs elsewhere.

Why league tables of cost-per-QALY figures mislead

A league table ranks interventions by their cost per QALY or cost per life year gained, usually with figures taken from separate published studies, and its entries may not be comparable. Drummond, Torrance and Mason traced the entries of a published league table back to their source studies and found differences in discount rate, in the method used to value health states, in the range of costs and consequences counted and in the comparison programme; a related BMJ paper by the same authors also noted differences in year of origin and setting.

The worked example adds a further problem. A table that places an average ratio beside an incremental one, or ratios calculated against different comparators, sets side by side figures that answer different questions, since treatment B alone could enter a table at £3,235, £22,500 or £40,000 per QALY. A league table also lists options for the same patients as if they were independent, when only a fully incremental comparison among them is valid. The authors did not reject league tables outright. They concluded that a table is useful only when decision makers can judge the relevance and reliability of the evidence for their own setting, called for fuller reporting of methods and results, and argued that without systematic comparisons value for money would probably be compared informally, with a worse balance of risk and benefit.

Reading a cost per QALY against NICE's £25,000 to £35,000 range

A cost per QALY becomes a decision input only when it is compared with a cost-effectiveness threshold. NICE's manual (PMG36) states that the committee does not use a precise maximum acceptable ICER, and that the appropriate maximum reflects the opportunity cost of programmes displaced by new, more costly technologies. Because NICE lacks complete information about the costs and QALYs of all competing programmes, it uses a range rather than a single figure. The manual was updated in March 2026 to include the change to that range, after a government announcement on 1 December 2025 that the previous range of £20,000 to £30,000 per QALY gained would rise to £25,000 to £35,000 from April 2026.

Under section 6.3 of the manual, the committee needs increasingly strong reasons to recommend a technology as its most plausible ratio rises through and above £25,000 to £35,000 per QALY gained, applies the same range to cheaper and less effective technologies, and asks for net health benefits at both values; the threshold page sets out these rules in full. In the worked example, B's incremental net health benefit over A is about −0.04 QALYs even at £35,000, the same conclusion as the £40,000 ratio expressed as health lost.

Discounting, perspective and the reference case behind a reported figure

A reference-case (base-case) cost per QALY submitted to NICE is calculated under the reference case in PMG36, and each element of it changes the number. Two figures for the same technology can therefore differ because of conventions rather than evidence.

The perspective decides which costs enter the numerator. The manual allows a non-reference-case discount rate of 1.5% a year for both costs and health effects only when the technology is for people who would otherwise die or have a very severely impaired life, is likely to restore them to full or near-full health, and has benefits likely to be sustained over a very long period.

Discounting matters most when costs fall early and QALY gains fall late. A QALY gain in year $t$ is discounted as follows.

$$Q^{*} = \frac{Q_t}{(1 + r)^{t}}$$

where $Q^{*}$ is the present value of the gain, $Q_t$ is the undiscounted gain in year $t$ and $r$ is the annual discount rate. In an illustrative case, a treatment costs £20,000 more than its comparator, all spent now, and yields one extra QALY in year 10. At 3.5% the discounted gain is $(1.035)^{-10} \approx 0.70892$ QALYs, so the ratio is as follows.

$$\frac{20000}{1 / 1.035^{10}} = \frac{20000}{0.70892} \approx 28212$$

where the result is the cost in pounds per discounted QALY gained. At 1.5% the discounted gain is $(1.015)^{-10} \approx 0.86167$ QALYs and the ratio falls to about £23,211. The rate alone moves this treatment from above £25,000 to below it, so a cost per QALY has to be read with the rate used to produce it.

Negative ratios and the south-west quadrant

Dividing two differences loses the information about which of them is negative, so the sign of a cost per QALY cannot be read without the two differences behind it. The illustrative cases below use the same cost difference of £2,000 and QALY difference of 0.10 in absolute size, with signs varied; the ICER page explains the quadrants in full.

Misreadings of a cost per QALY

Most misreadings treat a ratio between two options as a fixed property of one of them, or treat a single estimate as certain. The ratio's limits show up in three ways.

A cost per QALY describes a comparison between two options rather than a price attached to the treatment. It changes when the comparator, population, time horizon or discount rate changes, even if the treatment's price does not. At the margin it does say what extra health costs. In the worked example, funding B over A buys QALYs at £40,000 each.

A low cost per QALY does not show that a technology is affordable, because the ratio says nothing about how many people will be treated. PMG36 handles affordability separately. When a medicine's potential net budget impact is expected to exceed £40 million a year in any of its first 3 financial years of NHS use, NHS England and NHS Improvement will offer commercial discussions with the company before requesting a variation to the funding requirement. Budget impact analysis addresses that question.

A single figure also hides uncertainty. PMG36 asks models to quantify decision uncertainty, the probability that a different decision would be reached if the true cost effectiveness were known.

Sources

• Drummond MF, Sculpher MJ, Claxton K, Stoddart GL, Torrance GW. Methods for the Economic Evaluation of Health Care Programmes. 4th ed. Oxford: Oxford University Press; 2015.
• Drummond M, Torrance G, Mason J. Cost-effectiveness league tables: more harm than good? Social Science and Medicine. 1993;37(1):33-40. doi:10.1016/0277-9536(93)90315-u
• Mason J, Drummond M, Torrance G. Some guidelines on the use of cost effectiveness league tables. BMJ. 1993;306(6877):570-572. doi:10.1136/bmj.306.6877.570
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; published 31 January 2022, last updated 31 March 2026. Sections 4.2.7, 4.2.9, 4.2.15, 4.2.16, 4.5.1 to 4.5.3, 4.7.2, 4.10.8, table 4.1, 5.10.4 to 5.10.6, 6.3.1 to 6.3.9 and Update information.
• National Institute for Health and Care Excellence. Changes to NICE's cost-effectiveness thresholds confirmed. News article, 1 December 2025.