Article Text

Download PDFPDF

Health economic methodology illustrated with recent work on Chlamydia screening: the concept of extended dominance
  1. M J Postma1,
  2. R de Vries1,
  3. R Welte2,
  4. W J Edmunds3
  1. 1
    Groningen Research Institute of Pharmacy (GRIP), University of Groningen, Groningen, The Netherlands
  2. 2
    Helmholtz Center Munich, German Research Center for Environmental Health, Institute of Health Economics and Health Care Management, Neuherberg, Germany
  3. 3
    Health Protection Agency, Communicable Disease Surveillance Centre, London, UK
  1. Professor M J Postma, Groningen Research Institute of Pharmacy (GRIP), University of Groningen, Antonius Deusinglaan 1, 9713 AV, Groningen, The Netherlands; m.j.postma{at}


  • Competing interests: None declared.

View Full Text

Statistics from

Two economic evaluations were recently published on screening for Chlamydia trachomatis in The Netherlands and the United Kingdom.1 2 Both papers analysed various alternative screening strategies, the Dutch paper regarding different screening intervals, the British paper targeting different groups regarding age, gender and sexual behaviour. Besides presenting formal cost-effectiveness ratios in net costs per quality-adjusted life-year (QALY) gained, both papers additionally analysed results using the health-economic concept of dominance. This brief report aims to make the readers of this journal acquainted with this concept and the related issue of extended dominance. Extended dominance is covered as a theoretical topic in many health-economic text books, but is only scarcely encountered in daily practice. Both papers mentioned on Chlamydia screening do, however, nicely illustrate the concept in all its dimensions. For clarity, we have added a box with the health-economic terms used in this paper.3


De Vries et al1 analysed a systematic screening programme for men and women with various screening frequencies: one-off, any 10, any five, any two and any year. The paper builds on previous work published since 2000.46 Calculations were performed for a population of 50 000 men and 50 000 women aged 15–29 years. Previous research demonstrated that the exact risk of pelvic inflammatory disease (PID) as a major complication of Chlamydia infection was crucial for the outcome of such cost-effectiveness models. In this analysis, the risk of PID was put at 20% per infected case. The main findings were that one-off screening would be cost-saving and that up to a two-year frequency screening could be considered cost-effective.

Adams et al2 recently published options for Chlamydia screening in the United Kingdom in this journal. The population investigated comprised 40 000 individuals aged 16–45 years, again equally divided between both sexes.7 8 Annual screening was compared for various age cut-offs, for all women, selectively for only those women who recently changed sex partner(s), and for all women and men annually. PID risk was varied over three values: 1%, 10% and 30%. The main findings were that the cost-effectiveness of screening was sensitive to the proportion of infections that go on to develop PID, that screening all young women below 20 years of age was always most cost-effective and that the strategy of targeting both young men and women could still be better than many other options investigated.

Important differences between the two papers concern the timing in the models and the size and age distributions of the populations analysed. In particular, de Vries et al1 analysed one-off screening (and costs) in relation to cost savings and QALY gained during 20 subsequent years, whereas Adams et al2 analysed annual screening (and costs) in relation to those savings and QALY for 10 years. The differences in timing, as well as in population sizes and model structures, present explanations for order-of-magnitude differences in the results of both studies. Important similarities between the two papers concern the use of dynamic models and a rigid incremental approach. Dynamic models are used to describe the transmission dynamics of Chlamydia and so include the secondary epidemiological and health-economic effects of screening. Appropriately parameterised dynamic models are seen as providing the most reliable estimates for the cost-effectiveness of Chlamydia screening in the general population.9 We formally introduce the incremental approach in the Methods section as one health-economic term considered.


In this section, we briefly review the health-economic terminology used in this paper (see box). Our use of the term cost-effectiveness analysis also comprises cost-utility analysis. In particular, cost-utility analysis relates differences in costs of alternatives to differences in QALY, whereas the broader cost-effectiveness analysis may also consider other health outcomes, such as PID and infections. For the calculation of the incremental cost-effectiveness ratio (ICER), strategies are ranked in order of costs and subsequently compared two-by-two.

Dominance refers to the situation in which one alternative is dominated by another, and the dominated alternative should be ruled out of consideration according to stringent health-economic reasoning (see box). Health economists distinguish between strict and extended dominance (also referred to as strong and weak, respectively). They formally define the strictly dominant strategy as one that is cheaper and renders greater health gains than the alternative. As an expansion on the concept, health economists formally define extended dominance as the situation in which the index strategy is not dominated by one other, but rather by a linear combination of two other alternatives.10 The argument being that by combining the two alternatives wisely in a mixed strategy, a new strategy will result that strictly dominates the index strategy.


De Vries et al1 indeed found in their research that two specific alternatives dominate a third alternative, defining the situation of extended dominance. No strict dominance was found: no single individual strategy dominated another in terms of both lower costs and more effects.1 Both one-off screening (A) and screening every 5 years (C) dominated screening every 10 years (B; see table 1). So, according to formal health-economic decision rules, the authors removed alternative B from further analysis because of extended dominance. Providing one-off screening for only 50% of the population and repeated screening after five years for the other half (D) provided lower costs and more QALY than screening the whole population every 10 years (the 50-50 distribution elaborated here is not necessarily the adequate distribution in other cases; generally a range of distributions applies that may or may not include the 50-50 option). Also, table 1 shows that the value for money, in terms of a lower ICER, was better for C than for B. A lower ICER for a more expensive strategy compared with a previous strategy is another way of identifying extended dominance (see box).

Table 1 Extra costs and QALY gained over 20 (de Vries et al1) and 10 (Adams et al2) year periods for two studies on Chlamydia screening within several strategies

Health-economic terms used in this paper (partly taken and adapted from Berger et al3)

Cost-effectiveness analysis: This is a systematic method of comparing two or more alternative programmes by measuring the costs and consequences of each. A distinguishing feature of cost-effectiveness analysis is that the consequences (health outcomes) of all the programmes to be compared must be measured in the same common units—natural units related to the clinical objectives of the programmes (eg symptom-free days gained, cases prevented, patients improved, life-years gained). Also QALY may be the outcome, and then the cost-effectiveness may be further specified as being a cost-utility analysis (cost-utility analysis being conceived as a subform of cost-effectiveness analysis).

Cost-effective: A strategy is labelled “cost-effective” if the ICER is below a specified threshold for cost-effectiveness, this may for example be £30 000 per QALY.

Cost-effectiveness plane: The cost-effectiveness plane results from plotting the difference in effects of two alternatives (eg no screening and screening) versus the respective difference in costs. Typically, effects are measured on the x-axis and costs on the y-axis (see fig 1).

Figure 1 Cost-effectiveness plane for alternatives E to G (diamonds) from Adams et al 2 and the combined strategy H (see also table 1). Extra costs and quality-adjusted life-years (QALY) are measured compared with no screening. F (selective screening) is extended dominated by E (women only) and G (men and women), as we can find a piece of the line in between E and G (bold) on which any option strictly dominates F (ranging from equal QALY and fewer costs than F on the left-hand side, to equal costs and more QALY than F on the right-hand side of the bold line). The position of H from table 1 (both lower costs and more effects) is grossly indicated on the bold line piece. Note that for clarity of representation the x-axis starts at 50 QALY.

Cost saving: A strategy is labelled “cost saving” if its downstream savings on healthcare resource use exceeds the costs invested in the strategy.

Dominance: This refers to a situation in which one alternative is dominated by another, and accordingly the dominated alternative should be ruled out of contention, according to strict health-economic reasoning. There are two ways an alternative can be dominated: strict and extended. In strict dominance, there is another alternative that is both more effective and less costly. In extended dominance, there is another alternative that is more effective and more costly, but provides better value for money (lower ICER). Also, with extended dominance, two alternatives can be selected that when combined provide strict dominance regarding a third alternative.

ICER (incremental cost-effectiveness ratio): This refers to the ratio of the difference in costs (incremental costs) divided by the difference in outcomes (incremental effect) between two alternative programmes. If there are more than two alternatives, programmes are compared on a systematic pair-wise basis using their ICER.

QALY (quality-adjusted life-year): The QALY is a universal health outcome measure applicable to all individuals and all diseases, thereby enabling comparisons across diseases and across programmes. The QALY combines, in a single measure, gains or losses in both quantity of life (mortality) and quality of life (morbidity).

Strong and simple dominance: These are synonyms for strict dominance.

Weak dominance: This is a synonym for extended dominance.

Besides strict dominance, Adams et al2 also found extended dominance. In particular, the authors reported on the strategy of selectively targeting women below 20 years of age and a recent change in sex partner(s), a strategy being extended dominated by the two other strategies targeted at those aged <20 years (table 1; for illustrative purposes, we present the scenario for a 30% PID risk). Besides strategies E (women only), F (selective) and G (men and women), table 1 shows the combined strategy H. Fig 1 shows all strategies within the formal cost-effectiveness plane (see box for definition of this plane). Stringent health-economic reasoning implies the exclusion of selective screening (F) from the set of policies considered for implementation, limiting the set of alternatives to E and G. Then, based on the acceptability of the ICER, either all women should be screened (E) or, if still acceptable, the strategy targeting both men and women (G) should be adopted. Corresponding calculation of the ICER should be from E to G, rather than from E to F, as the latter has been ruled out.10 Also, this ICER at £8178 per QALY, exactly reflects the cost-effectiveness ratio for adding men to the screening programme yet directed towards women only. Note that this value also represents the ICER for our previously defined combined strategy H.


We note that extended dominance is a theoretical concept and that for various reasons combining two strategies is practically difficult. For example, in the analysis of de Vries et al1 inequalities are obviously introduced if parts of the population would be screened at a different interval than other parts. Also, alternatives normally excluded through extended dominance may be preferred for reasons beyond health-economics. For example, a strategy may be chosen based on ease of implementation, acceptability to the populations concerned, or political attractiveness. So, in practice alternatives excluded through extended dominance may still be considered in policy making, despite their inferiority from a formal health-economic standpoint. This presents a good reason to report on such strategies in the same detail as on non-dominated alternatives. Adams et al.2 did not exclude alternatives as a result of extended dominance as some of them were strategies explicitly considered by the National Chlamydia Screening Programme. Given this political attention, these strategies warranted careful elaboration.


We reviewed one specific concept in health-economic theory of extended dominance on the basis of two recently published studies on Chlamydia screening cost-effectiveness, one of them in this journal. The concept is often applied to narrow the set of options from which to choose. We emphasise that it is important always to perform such a dominance analysis. For assessing extended dominance, the options defined as new strategies are combinations of alternatives: provide part of the population with one alternative and the remaining part with another alternative. These combinations obviously pose better options from a purely health-economic perspective. Although this search for health-economic optima is tempting, combined strategies may be difficult to implement in practice, because of equity concerns. Also, extended dominance may lead to the exclusion of policy options that are still attractive for other reasons than health economics. Therefore, we recommend always to report fully on such extended dominated alternatives, in order to provide decision makers with the full picture on all the possible alternatives being discussed.


View Abstract


  • Competing interests: None declared.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.