trial and, if an intermediate end‐point shows significant benefit from the new intervention, the trial may be stopped early before mature survival data can be obtained. Under these circumstances, some mathematical modelling of the likely survival benefit may be required. Extrapolation always introduces an element of uncertainty into the estimate of survival gain.
Deriving the utility values is more controversial and requires some assessment of the relative desirability of various health states. Sometimes the value attached to each health state is taken from patients with actual experience of the health state but often this is not possible and so values taken from health professionals with knowledge of the health state, or members of the public to whom the health state has been described, may be used. Indeed the question of whether the perspective of the patient experiencing the health state or the perspective of the public (payers in the healthcare system) is the more important is one which remains controversial.
Often the value of the health state is obtained directly using one of three common techniques:
1 Visual Analogue Scale – respondents are asked to mark a line (usually labelled 0–100) with an ‘X’ to indicate their valuation of the health state. The distance from 0 to X is measured and divided by 100 to give the utility value. This is a simple technique but has the disadvantage that the respondent is not asked to make any sort of trade‐off or sacrifice in valuing the health state. It is therefore considered an inferior technique by economists.
2 Time Trade‐Off – in this technique, individuals are asked to choose between living for a given time in a state of poor health and living for a shorter time in full health. For example, respondents are asked to imagine they have 10 years left to live and they will live these 10 years in the health state being valued. They are then asked to decide how many years they would be willing to give up in order to have full health. If a respondent indicates they would be willing to give up 2 of the 10 years to be ‘cured’, this would translate into a utility value of 0.8. A worse health state, in which a respondent would give up 5 years to be free of the condition, would result in a utility value of 0.5.
3 Standard Gamble – respondents are asked to imagine being in the health state being valued and are then told that there is a treatment which could restore them to full health; however, the treatment also has the possibility of causing immediate death. They are asked to indicate what risk of death they would be willing to accept in order to be ‘cured’, a willingness to accept a higher risk of instant death leading to the value on the current health state being lower. Once again survival is being sacrificed (or at least risked) to improve quality of life.
An alternative approach to deriving utility values is an indirect method where a questionnaire is used to describe the current health state in terms of patient experience and function and then the results from the questionnaire are converted into a utility value by inviting patients or the general public to value the health states by the methods above. The EQ‐5D questionnaire is commonly used and it assesses quality of life in terms of five key dimensions; mobility, self‐care, usual activities, pain/discomfort and anxiety/depression. Each of these dimensions can be rated on either a three‐point (EQ‐5D‐3L) or five‐point (EQ‐5D‐5L) with the latter version being developed to overcome the potential for reduced sensitivity to changes in health state in the three‐level version. The resulting potential health states from the EQ‐5D have been converted to utility values using surveys of members of the general public in the UK.
None of the available methods of deriving utility values is perfect but used individually or in combination some meaningful assessment of quality of life can be obtained to factor into QALY calculation. Table 3.1 shows examples of utility values for a number of common health states.
Table 3.1 Utility values from selected published studies.
| Disease state | Utility value |
|---|---|
| Acute coronary syndrome (Karnon et al. 2008) | 0.80 |
| Progressive glioma (Rogers et al. 2008) | 0.73 |
| Relapse in multiple sclerosis (Gani et al. 2008) | 0.50 |
| Mild to moderate depression (Kendrick et al. 2009) | 0.70 |
| Symptomatic ulcer (Latimer et al. 2009) | 0.55 |
How is cost‐effectiveness then assessed?
As noted above, the convention in health economics is to describe the extra health gain from the more beneficial treatment and then to define the cost of this extra benefit. In CUA, this leads to a measure known as the incremental cost‐effectiveness ratio (ICER), which is described in Figure 3.2.
Figure 3.2 Incremental cost‐effectiveness ratio.
From the outcome of the simple calculation in Figure 3.2, one can give a value for the ICER as the extra cost per QALY gained, commonly known as ‘cost per QALY’.
To help decide if a treatment is cost‐effective, the resulting ‘cost per QALY’ can be compared with ‘costs per QALY’ for other interventions already adopted by the health service or to the decision maker's stated ‘willingness‐to‐pay’ for an additional QALY. In the UK, for many of the medicines it appraises, NICE uses a rule of thumb or ‘threshold’ that a cost per QALY of less than £20 000 is generally considered to be cost‐effective while a cost per QALY greater than £30 000 is generally considered not to be cost‐effective; decision makers in other countries may adopt other willingness‐to‐pay thresholds. Table 3.2 shows the ICERs of some common healthcare interventions.
Table 3.2 Cost per QALY gained figures from selected published studies (references at end of chapter).
| Treatment | Cost per QALY gained |
|---|---|
| Clopidogrel in acute coronary syndrome | £2 284 |
| Carmustine wafers in glioma | £54 500 |
| Natalizumab in multiple sclerosis | £18 700 |
| SSRIs in mild to moderate depression | £14 854 |
| Addition of a proton pump inhibitor to NSAID or COX 2 selective inhibitors | £1 000 |
SSRI, selective serotonin reuptake inhibitor; NSAID, non‐steroidal anti‐inflammatory drug.
A simple example showing how such QALYs are calculated is shown in Table 3.3.
