SOCIETAL VIEWS ON NICE, CANCER DRUGS FUND AND VALUE-BASED PRICING CRITERIA FOR PRIORITISING MEDICINES: A CROSS-SECTIONAL SURVEY OF 4118 ADULTS IN GREAT BRITAIN

Authors

  • Warren G. Linley,

    1. Centre for Health Economics & Medicines Evaluation, Institute of Medical and Social Care Research, Bangor University, Bangor, UK
    Search for more papers by this author
  • Dyfrig A. Hughes

    Corresponding author
    • Centre for Health Economics & Medicines Evaluation, Institute of Medical and Social Care Research, Bangor University, Bangor, UK
    Search for more papers by this author

Correspondence to: Centre for Health Economics & Medicines Evaluation, Institute of Medical and Social Care Research, Bangor University, Dean Street, Bangor, Gwynedd, UK. E-mail: d.a.hughes@bangor.ac.uk

ABSTRACT

The criteria used by the National Institute for Health and Clinical Excellence (NICE) for accepting higher incremental cost-effectiveness ratios for some medicines over others, and the recent introduction of the Cancer Drugs Fund (CDF) in England, are assumed to reflect societal preferences for National Health Service resource allocation. Robust empirical evidence to this effect is lacking. To explore societal preferences for these and other criteria, including those proposed for rewarding new medicines under the future value-based pricing (VBP) system, we conducted a choice-based experiment in 4118 UK adults via web-based surveys. Preferences were determined by asking respondents to allocate fixed funds between different patient and disease types reflecting nine specific prioritisation criteria. Respondents supported the criteria proposed under the VBP system (for severe diseases, address unmet needs, are innovative—provided they offered substantial health benefits, and have wider societal benefits) but did not support the end-of-life premium or the prioritisation of children or disadvantaged populations as specified by NICE, nor the special funding status for treatments of rare diseases, nor the CDF. Policies introduced on the basis of perceived—and not actual—societal values may lead to inappropriate resource allocation decisions with the potential for significant population health and economic consequences. Copyright © 2012 John Wiley & Sons, Ltd.

1 INTRODUCTION

The UK National Health Service (NHS) has legal and moral obligations to provide fair, comprehensive, needs-based care for all [Department of Health, 2010a]. Given the unprecedented efficiency savings demanded across the NHS in recent and coming years [Department of Health, 2009; Institute for Fiscal Studies/Nuffield Trust, 2012], it is imperative that resource allocation decisions provide the most effective and sustainable use of finite resources. The National Institute for Health and Clinical Excellence (NICE) makes compulsory recommendations on the use of medicines and other health technologies in the NHS in England and Wales, with reference to their clinical and cost effectiveness. The funding of new medicines requires that other existing medicines or services are displaced, the opportunity cost of which is reflected in NICE's cost-effectiveness threshold, set at £20,000–£30,000 per quality-adjusted life-year (QALY) gained [National Institute for Health and Clinical Excellence, 2008a]. However, several medicines with incremental cost-effectiveness estimates in excess of this threshold range have been approved by NICE for use via the NHS (e.g., sunitinib for advanced renal cancer and riluzole for motor neurone disease) [Rawlins et al., 2010].

Justification for this departure from the usual cost-effectiveness threshold range includes the social value judgements of NICE's Citizen Council. On the basis of its views, six specific criteria besides clinical and cost-effectiveness have recently been put forward as reflecting societal preferences in the allocation of health resources [Rawlins et al., 2010] (Table 1). Despite these laudable efforts to incorporate societal views into the NICE work programme, the extent to which this group of 30 lay persons can reflect the views and preferences of the public as a whole regarding the allocation of scarce health resources has been questioned [Buxton and Chambers, 2011], and it is suggested that access to some new medicines, such as those to treat cancer, may still be inappropriately restricted [Department of Health, 2010b].

Table 1. Current and proposed medicines prioritisation criteria explored
Prioritisation criteria exploredReported rationale for use of criteria within UK NHSScenario construct
  1. AGNSS, Advisory Group for National Specialised Services; AWMSG, All Wales Medicines Strategy Group; CDF, Cancer Drugs Fund; Dept Health, Department of Health; NHS, National Health Service; NICE, National Institute for Health and Clinical Excellence; SMC, Scottish Medicines Consortium; VBP, Value-Based Pricing.

Severity of underlying diseaseNICE: society would generally give priority to the expensive relief of a very serious condition than to the inexpensive relief of a mild condition [Rawlins et al., 2010].All else being equal: medicine for severe disease compared against medicine for moderate disease, as mild disease may be viewed as not necessitating treatment (in cohorts 1 and 2).
VBP: society may place a greater weight on treating particularly severe or life threatening conditions [Dept Health, 2010d].Trade-off scenarios: as severe disease is the criterion in question, trade-off scenarios explored smaller health gains (in cohort 1) and higher costs for treatment (in cohort 2) of severe disease compared with moderate disease.
Unmet needVBP: the current system [of appraisal] may not fully reflect society's preferences if there are no existing alternative treatments and so a significant unmet need [Dept Health, 2010d].All else being equal: medicine for disease with several other treatments available via the NHS compared against medicine for a disease with only one medicine available (in cohorts 1 and 2).
 Trade-off scenarios: as unmet need is the criterion in question, trade-off scenarios explored smaller health gains (in cohort 1) and higher costs for treatment (in cohort 2) for the disease with only one medicine available compared with the disease with several treatments available.
Significant innovationNICE: product produces a demonstrable and distinct benefit of a substantial nature that may not be adequately captured in the quality of life measure used [Rawlins et al., 2010].All else being equal: medicine that works in a new way compared against medicine that works in similar way to existing alternatives for treatment of the same disease (in cohorts 1 and 2).
VBP: a treatment representing a significant breakthrough and an important advance over existing therapies would provide a large QALY benefit. It could also be represented by a qualitative assessment of the innovation reported by a new medicine reflecting, for example, new modes of action [Dept Health, 2010d].Trade-off scenarios: as innovation should deliver an advantage, the only plausible trade-off scenarios are an improvement in health for the medicine that works in a new way (in cohort 1) and an improvement in health for the medicine that works in a new way accompanied by an additional cost for the medicine that works in a new way (in cohort 2).
Wider societal benefitsVBP: impacts of a product beyond direct health effects. These might include benefits related to reduced reliance on carers and other wider societal factors [Dept Health, 2010d].All else being equal: medicine for disease that causes patients to be reliant upon carers (e.g. family members) for day-to-day needs and reduces that reliance on carers, compared against medicine for disease that does not cause patients to be reliant upon carers (in cohort 2 only).
 Trade-off scenario: medicine for disease that causes patients to be reliant upon carers (e.g. family members) for day-to-day needs, and reduces that reliance on carers, is more costly compared with medicine for disease that does not causes patients to be reliant upon carers (in cohort 2 only).
Disadvantaged populationsNICE: the NHS gives special priority to improving the health of the most disadvantaged members of the population, particularly poorer people and ethnic minorities [Rawlins et al., 2010].All else being equal: medicine for treatment of disease that typically affects disadvantaged populations (e.g. those form low income families) compared against medicine for disease that does not typically affect disadvantaged populations (in cohorts 1 and 2).
 Trade-off scenarios: as disadvantaged populations are the criterion in question, trade-off scenarios explored smaller health gains (in cohort 1) and higher costs for treatment (in cohort 2) of disadvantaged populations compared with non-disadvantaged populations.
ChildrenNICE: compilation of evidence and assessment of improvements in the quality of life in children are methodologically challenging. Society would generally favour ‘the benefit of the doubt’ being afforded to sick children [Rawlins et al., 2010].All else being equal: medicine for treatment of children compared against medicine for treatment of adults (in cohorts 1 and 2).
 Trade-off scenarios: as children are the criterion in question, trade-off scenarios explored smaller health gains (in cohort 1) and higher costs for treatment (in cohort 2) for children compared with adults.
End-of-life treatmentsNICE: the public generally places special value on treatments that prolong life—even for a few months—at the end of life as long as that extension of life is of reasonable quality [Rawlins et al., 2010]. The end-of-life policy specifies that patients should have a short life expectancy, normally of less than 2 years, and the gain in life expectancy over currently available NHS treatment should normally exceed 3 months [NICE, 2009].All else being equal: medicine for treatment of fatal disease that leads to death in 18 months without treatment compared against medicine for treatment of fatal disease that leads to death in 60 months without treatment. Both medicines extend life by 6 months (in cohorts 1 and 2).
Trade-off scenarios: as patients meeting NICE's end-of-life policy reflect the criterion in question, trade-off scenarios explored smaller life extension of 3 months (in cohort 1) and higher costs for treatment (in cohort 2) for patients with life expectancy of 18 months without treatment compared with patients with life expectancy of 60 months without treatment.
Cancer treatmentsCDF: it is possible that society values health benefits to patients with cancer more highly, all else being equal, than benefits to patients suffering other conditions [Dept Health, 2010b].All else being equal: medicine for treatment of potentially fatal cancer compared against medicine for treatment of potentially fatal non-cancer disease (in cohorts 1 and 2).
Trade-off scenarios: as cancer is the criterion in question, trade-off scenarios explored smaller health gains (in cohort 1) and higher costs of treatment (in cohort 2) for patients with cancer compared with patients with non-cancer disease.
Rare diseasesAGNSS: top-sliced funds for treatments of exceptionally rare diseases in England [NHS Specialised Services, 2011].All else being equal: medicine for treatment of rare disease compared against medicine for treatment of common disease (in cohorts 1 and 2).
SMC: policy for appraising orphan drugs [SMC, 2010].Trade-off scenarios: as rarity is the criterion in question, trade-off scenarios explored smaller health gains (in cohort 1) and higher costs of treatment (in cohort 2) for rare disease compared with common disease.
AWMSG: policy of appraising ultra-orphan drugs [AWMSG, 2011].
Stakeholder persuasionNICE: patients and their advocates…can explain where symptomatology of their condition is poorly reflected in clinical trials and health-related quality of life measure [Rawlins et al., 2010].No practical scenario construct possible. Not explored in this study.

The Cancer Drugs Fund (CDF) was introduced in England in 2011 to facilitate access specifically to cancer medicines that have received a negative opinion from NICE on the grounds that they do not represent a good use of NHS resources, or that have not yet been appraised. The government justified the CDF—set at £200m per annum—on the basis that: “…it is possible that society values health benefits to patients with cancer more highly, all else being equal, than benefits to patients suffering other conditions” [Department of Health, 2010b]. Although disease severity is consistently viewed as a valid criterion for prioritising health resources [Dolan et al., 2005; Shah, 2009], we are unaware of any empirical evidence for the preferential funding of cancer treatments.

From 2014, all new branded medicines in the UK will be priced according to their therapeutic value and wider benefits they may deliver [Department of Health, 2010c; Department of Health, 2011]. Under this value-based pricing (VBP) system, it is proposed that explicit weightings be attached to the health benefits (QALY gains) provided by medicines to reflect a broader range of relevant criteria [Department of Health, 2010d] (Table 1). Again, with the exception of severity of disease, empirical evidence of the desirability of these criteria for rewarding new medicines with premium prices seems lacking [Department of Health, 2010c].

It is apparent, therefore, that current prioritisation criteria used in pricing and reimbursement systems in the UK, and recent initiatives to address their perceived short-comings, are without robust supporting empirical evidence that they reflect societal preferences. This may lead to inappropriate resource allocation decisions, which take on a greater importance in the context of the increasing financial pressures under which the NHS is operating.

Our study explored societal preferences for the prioritisation criteria used by NICE, those proposed under the VBP system, and the UK government's assumptions used to justify the introduction of the CDF. In addition, we explored whether a societal preference exists for treating rare diseases over more common diseases, given that funds are top-sliced for certain treatments of very rare diseases in England [NHS Specialised Services, 2011], and both the All Wales Medicines Strategy Group (AWMSG) and the Scottish Medicine Consortium (SMC) permit additional considerations in the appraisal of medicines for the treatment of such diseases [All Wales Medicines Strategy Group, 2011; Scottish Medicines Consortium, 2010].

2 METHODS

2.1 Questionnaire design

We reviewed relevant documents and policies [All Wales Medicines Strategy Group, 2011; Department of Health, 2010b; Department of Health, 2010c; Department of Health, 2010d; Department of Health, 2010e; Department of Health, 2011; NHS Specialised Services, 2011; National Institute for Health and Clinical Excellence, 2008a; National Institute for Health and Clinical Excellence, 2008b; National Institute for Health and Clinical Excellence, 2008c; National Institute for Health and Clinical Excellence, 2009; Rawlins et al., 2010; Scottish Medicines Consortium, 2010] to identify nine specific prioritisation criteria (besides clinical-effectiveness and cost-effectiveness) for exploration within our study (Table 1).

We used a choice-based format in which adult members of the general public were asked to express their preferred way for the NHS to allocate resources between two competing hypothetical populations. Respondents selected one of the 11 alternative resource configurations ranging from all money to be spent on one population, through an equal distribution, to all money to be spent on the alternative population, as illustrated in Figure 1 using disease severity as an example criterion (Web Appendix I for further details).

Figure 1.

Summary of questionnaire format using disease severity as an example criterion

The descriptions of the populations and their treatments were constructed to isolate as far as possible the influence of the criterion in question. We initially constructed a single questionnaire consisting of three-part questions: a scenario of all else being equal and two subsequent trade-offs. This was piloted amongst a convenience sample of 23 adults with a broad range of educational and occupational backgrounds. None of the pilot respondents reported difficulties in understanding the question framing, terminology or task required; however, to reduce respondent burden and completion time, the questionnaire was subsequently divided into two versions.

Each question of the final versions of the questionnaires consisted of two parts. Part 1, common to both questionnaires, represented a scenario of ‘all else being equal’ in which only the criterion in question differed between the competing populations; the costs and effectiveness of treatment and all other aspects of the underlying condition were identical. Part 2 differed between the two versions of the questionnaires and was included to test if any preferences for the criterion under a scenario of ‘all else being equal’ were retained under less favourable effectiveness and/or cost conditions. In cohort 1, we explored preferences when faced with a trade-off in total health benefits but retained the assumption of equal costs, whereas in cohort 2, we explored preferences when faced with a twofold change in costs, which in the context of a fixed NHS budget represents a trade-off in the total number of patients who could be treated.

2.2 Administration

We commissioned Vision Critical Research Solutions (UK) Ltd to administer the two web-based questionnaires simultaneously to a broadly UK-representative sample of its active survey panel based in England, Scotland, and Wales in August 2011. There are no formal methods of sample size calculation for this type of study. We therefore determined our target sample size by reference to those reported in the empirical ethics literature [Dolan et al., 2005; Shah, 2009] and available resources. As the survey was closed when our target of 2000 complete responses was achieved in each independent cohort, it is not possible to determine a response rate.

The choice-based questions were presented to participants in random order to minimise the impact of ordering and learning effects across the cohorts [McColl et al., 2001]. An initial 100 panellists acted as an internal pilot to confirm that respondents were able to complete the questionnaires properly and within a reasonable timeframe before the survey invite was distributed more widely.

2.3 Analysis

Our primary null hypothesis was that, all else being equal, there would be no societal preferences for any of the criteria explored, that is, most respondents would prefer the NHS to divide resources equally between the competing populations. Secondary hypotheses were that, when faced with a trade-off in total health benefits, respondents in cohort 1 would prefer the NHS to fund treatment that resulted in greater overall health benefits, and when faced with a trade-off in costs, respondents in cohort 2 would prefer the NHS to fund the treatment that enabled most patients to be treated. Societal preferences were inferred from absolute majority responses.

Given the consistency of the findings across the two cohorts (Web Appendix II), responses to part 1 questions were meta-analysed using a conservative random effects model. Analyses of part 2 questions were conducted separately by categorising responses into three groups: respondents favouring either one of the two competing populations or respondents favouring an equal division between the two competing populations. Liddell exact test for matched pairs was used to determine the statistical significance of any relative shifts in preferences between both parts of each question.

Socio-demographic data were collected to assess the generalisability of the sample. Logistic regression modelling using age, health status, working status, country of residence and scenario-specific explanatory variables was conducted to determine their impact on respondents' expressed preferences. Analyses were performed in StatsDirect statistical software version 2.7.7, 2009 (StatsDirect Ltd, England).

3 RESULTS

A total of 4118 adults completed the questionnaires. Respondents' demographics were well-balanced across the two cohorts and were representatives of the population of Great Britain, with the exception of a lower proportion of respondents describing themselves as being in very good or good health and a higher proportion describing themselves as being in fair health. People aged 65 years and older were possibly under-represented and people aged 45–64 years slightly over-represented (Table 2). Residents of Northern Ireland, who represent less than 3% of the UK population, were not included amongst those surveyed.

Table 2. Socio-demographic characteristics of respondents and adult population of Great Britain
CharacteristicsCohort 1 (n, (%))Cohort 2 (n, (%))Great Britain (%)
  • *

    Figures for Great Britain based on adults aged 16 years and over in Office for National Statistics General Lifestyle Survey 2009 [Office for National Statistics, 2009].

  • Figures for Great Britain based on National Readership Survey 2010 population data [National Readership Survey, 2010].

  • Figure for Great Britain based on Office for National Statistics mid-2010 population estimates for adults aged 16 years and over [Office for National Statistics, 2011].

  • Δ

    Figures for Great Britain based on nomis official labour force statistics, seasonally adjusted percentage of people aged 16 years and over, June-August 2011 (covering our survey period) [Office for National Statistics, 2010].

  • People who are neither in employment nor unemployed (e.g. those who were looking after a home or retired).

Gender*
Male1026 (50.5)1000 (48.0)48.8
Female1007 (49.5)1085 (52.0)51.2
Age*
18–44 years926 (45.5)903 (43.3)48.8
45–64 years846 (41.6)900 (43.1)31.7
65 years and over261 (12.8)282 (13.5)19.5
Social grade
ABC11049 (51.6)1067 (51.2)55
C2DE984 (48.4)1018 (48.8)44
Working status‡, Δ
Employed1089 (53.6)1120 (53.7)58.0
Unemployed152 (7.5)144 (6.9)8.1
Economically inactive792 (38.9)821 (39.8)36.9
General health*
Very good/good1321 (65.0)1369 (65.7)79.0
Fair514 (25.3)536 (25.7)15.0
Bad/very bad198 (9.7)180 (8.6)6.0
Household composition*
With children582 (27.9)603 (29.7)25.0
Without children1503 (72.1)1430 (70.3)75.0
Household reliance on long-term informal care
Yes442 (21.7)435 (20.9)Unknown
No1591 (78.3)1650 (79.1)Unknown
Country
England1749 (86.0)1761 (84.5)86.4
Scotland186 (9.1)209 (10.0)8.6
Wales98 (4.8)115 (5.5)5.0

3.1 Preferences under assumption of ‘all else being equal’

Pooled responses to part 1 questions are presented in Table 3. All else being equal, a societal preference (based on an absolute majority) for allocation of NHS funds exists for treating patients with severe rather than moderate disease; for treating diseases where there are no alternative treatments available rather than diseases where several alternative treatment options exist; and for treating diseases that cause patients to be reliant upon carers rather than diseases that do not. For all other criteria, between 62% and 85% of respondents' allocations did not support a value premium.

Table 3. Preferences of respondents under assumption of all else being equal and when faced with trade-offs in health gains and costs
Scenario population 1ChoicePrioritise population 1Equal allocation to both populationsPrioritise population 2ChoiceScenario population 2
  % Respondents (95% CI)  
  • *

    Pooled results of cohorts 1 and 2 (n = 4118) using proportion meta-analysis, random effects model.

  • Reliance on carers explored in cohort 2 only, (n = 2085); RR = Relative risk point estimate based on Liddell exact test for matched pairs, used to compare the proportion of responses under trade-off conditions versus each cohort's responses to part 1 of each question.

  • Bold figures represent allocations with clear absolute majority and no overlap of confidence intervals; 95% CI = 95% confidence interval.

Severe diseaseAll else being equal*59.6 (58.1–61.1)31.0 (28.0–34.0)9.4 (6.0–13.5)All else being equal*Moderate severity disease
Little health improvement28.2 (26.2–30.2)42.6 (40.5–44.8)29.2 (27.2–31.2)Improves health considerably
RR = 0.12; p < 0.0001RR = 2.84; p < 0.0001RR = 5.76; p < 0.0001
Twice the cost of population 260.9 (58.8–63.0)30.2 (28.2–30.2)8.9 (7.7–10.2)Half the cost of population 1
RR = 1.12; p = 0.3193RR = 0.75; p = 0.0101RR = 1.53; p = 0.0176
 
No other medicines availableAll else being equal*56.5 (53.8–59.1)31.1 (28.9–33.1)12.5 (11.5–13.5)All else being equal*Several other medicines available
Little health improvement41.4 (39.3–43.6)36.3 (34.2–38.4)22.3 (20.5–24.2)Improves health considerably
RR = 0.22; p < 0.0001RR = 1.86; p < 0.0001RR = 3.11; p < 0.0001
Twice the cost of population 260.4 (58.3–62.5)27.2 (25.3–29.2)12.4 (11.0–13.9)Half the cost of population 1
RR = 2.22; p < 0.0001RR = 0.48; p < 0.0001RR = 0.93; p = 0.6753
 
Medicine works in a new wayAll else being equal*24.4 (22.1–26.9)62.2 (60.7–63.6)13.4 (12.3–14.5)All else being equal*Medicine works in similar way to existing medicines
Improves health considerably63.1 (60.9–65.2)29.1 (27.2–31.1)7.8 (6.7–9.1)Little health improvement
RR = 16.83; p < 0.0001RR = 0.10; p < 0.0001RR = 0.35; p < 0.0001
Improves health considerably and twice the cost of population 253.8 (51.6–56.0)32.8 (30.8–34.9)13.4 (11.9–13.9)Little health improvement and half the cost of population 1
RR = 8.98; p < 0.0001RR = 0.13; p < 0.0001RR = 0.93; p = 0.5399
 
Patients reliant on informal carersAll else being equal50.0 (47.8–52.1)40.6 (38.5–42.8)9.4 (8.2–10.7)All else being equalPatients not reliant on informal carers
Twice the cost of population 254.8 (52.7–57.0)34.3 (32.3–36.4)10.8 (9.5–12.3)Half the cost of population 1
RR = 1.57; p < 0.0001RR = 0.53; p < 0.0001RR = 1.34; p = 0.0443
 
Disadvantaged populationAll else being equal*34.5 (32.7–36.2)59.5 (57.6–61.5)6.0 (5.3–6.7)All else being equal*Not disadvantaged population
Little health improvement23.2 (21.3–25.1)47.9 (45.6–50.1)29.0 (27.0–31.0)Improves health considerably
RR = 0.33; p < 0.0001RR = 0.44; p < 0.0001RR = 14.43; p < 0.0001
Twice the cost of population 252.8 (50.6–54.9)38.5 (36.4–40.6)8.8 (7.6–10.1)Half the cost of population 1
RR = 6.33; p < 0.0001RR = 0.14; p < 0.0001RR = 2.27; p < 0.0001
 
ChildrenAll else being equal*37.5 (36.1–39.0)57.0 (55.5–58.5)5.5 (4.6–6.5)All else being equal*Adults
Little health improvement19.1 (17.4–20.9)44.5 (42.3–46.7)36.4 (34.4–38.6)Improves health considerably
RR = 0.15; p < 0.0001RR = 0.48; p < 0.0001RR = 20.97; p < 0.0001
Twice the cost of population 254.8 (52.6–56.9)38.3 (36.2–40.4)6.9 (5.9–8.1)Half the cost of population 1
RR = 4.86; p < 0.0001RR = 0.18; p < 0.0001RR = 1.93, p = 0.0005
 
18 months survival without treatmentAll else being equal*34.4 (30.4–38.6)47.6 (46.1–49.2)17.9 (15.5–20.5)All else being equal*60 months survival without treatment
3 month survival gain23.3 (21.5–25.2)50.5 (48.3–52.7)26.2 (24.3–28.2)6 month survival gain
RR = 0.30; p < 0.0001RR = 1.21; p = 0.0550RR = 2.3; p < 0.0001
Twice the cost of population 242.1 (40.0–44.2)39.2 (37.1–41.3)18.8 (17.1–20.5)Half the cost of population 1
RR = 1.78; p < 0.0001RR = 0.47; p < 0.0001RR = 1.41; p < 0.0001
 
CancerAll else being equal*30.8 (28.1–33.5)64.1 (61.5–66.7)5.1 (4.5–5.8)All else being equal*Non-cancer disease
Little health improvement20.8 (19.0–22.6)42.0 (39.8–44.1)37.3 (35.2–39.4)Improves health considerably
RR = 0.30; p < 0.0001RR = 0.23; p < 0.0001RR = 17.74; p < 0.0001
Twice the cost of population 247.5 (45.4–49.7)42.2 (40.1–44.4)10.3 (9.0–11.6)Half the cost of population 1
RR = 4.82; p < 0.0001RR = 0.15; p < 0.0001RR = 3.13; p < 0.0001
 
Rare diseaseAll else being equal*15.1 (14.0–16.2)43.2 (40.5–45.9)41.7 (38.2–45.3)All else being equal*Common disease
Little health improvement10.4 (9.1–11.8)32.4 (30.3–34.4)57.3 (55.1–59.4)Improves health considerably
RR = 0.45; p < 0.0001RR = 0.39; p < 0.0001RR = 5.54; p < 0.0001
Twice the cost of population 223.7 (21.9–25.6)38.0 (35.9–40.1)38.3 (36.2–40.4)Half the cost of population 1
RR = 3.00; p < 0.0001RR = 0.52; p < 0.0001RR = 0.82; p = 0.0784

3.2 Preferences under health gain trade-offs

Using each cohort's preferences under the assumption of all else being equal as a baseline, when faced with a trade-off in effectiveness, there was a statistically significant shift in preferences for all criteria towards the populations that gained a considerable improvement in health and away from the population that gained a little improvement in health (Liddell exact test p < 0.0001 in each case) (Table 3). A preference for treating diseases where there are no alternative treatments available remained present {proportion of respondents, (95% confidence interval [CI])} (41.4% [39.3%–43.6%]), despite the assumption of a little health gain in that patient group compared with a considerable health gain in patients with several treatment options available (22.3% [20.5%–24.2%]). A preference in favour of medicines that work in a new way was only apparent when coupled with a considerable improvement in health gains (63.1% [60.0%–65.2%]). Treatment of a common disease that produces considerable improvements in health gains was also strongly preferred (57.3% [55.1%–59.4%]) to treatment of a rare disease that produces a little improvement in health (10.4% [9.1%–11.8%]). There was no evidence of support of a value premium for any other criteria under effectiveness trade-off conditions.

3.3 Preferences under cost trade-offs

When faced with a trade-off in costs, there was a statistically significant shift in preferences for all criteria towards the populations that were more costly to treat (Liddell exact test p < 0.0001 in each case), with the exception of severity of disease (60.9% vs. 60.0%; relative risk [RR] = 1.12; p = 0.3193). This resulted in a significantly greater proportion of respondents expressing a preference for the most costly population than expressed a preference for either an equal division of resources or for the less costly population, with the exception of rarity of disease.

3.4 Impact of respondents' characteristics on preferences

Logistic regression analyses suggest that respondents' preferences are influenced by their individual characteristics and circumstances (Table 4). For example, those with children in their household were more likely to express a funding preference for treating children over adults than those without (odds ratio [OR] = 1.63 [95% CI 1.41–1.89]); those with a household reliance on carers were more likely to express a funding preference for medicines with wider societal benefits (e.g. reduced reliance upon carers) than those without (OR = 1.30 [95% CI 1.03–1.64]), and those in social grade C2DE were more likely to prioritise disadvantaged populations (e.g. those on low incomes) than those in social grade ABC1 (OR = 1.36 [95% CI 1.19–1.55]). Other observed funding preferences are less easy to explain; for example, compared with respondents rating themselves as in good/very good health, respondents rating themselves as in bad/very bad health were significantly less likely to favour the funding of medicines for severe diseases, medicines for conditions with no other treatment options, and medicines for children. There were no significant differences in preferences for any criterion based on country of residence.

Table 4. Logistic regression analyses under assumption of equal health gains and costs, odds ratios and 95% confidence interval
 Dependent variables—favoured versus (equal and not favoured)
Explanatory variablesSevere diseaseNo alternative treatment optionsMedicines work in new wayReliance on carersDisadvantaged populationsChildren18 months survivalCancerRare disease
  • Base values of explanatory variables:

  • Age: 18–44 years; general health: good/very good; working status: unemployed and economically inactive; country: England; children in household: no; reliance on carers: no; social grade: ABC1;

  • Each favoured dependent variable regressed against all general explanatory variables. Scenario-specific variables added and retained in respective models if provided a good fit based on Model X2 and/or deviance goodness of fit X2 and McFadden's pseudo-R2.

  • *

    Ideally Model X2 p < 0.05 and deviance goodness of fit X2 p > 0.05.

  • No evidence of collinearity among independent variables as assessed by tolerance statistics.

General explanatory variables—considered in all scenarios
Age: 45–64 years1.15 (1.01–1.32)1.16 (1.02–1.33)0.99 (0.85–1.16)1.23 (1.02–1.49)0.90 (0.78–1.04)1.05 (0.91–1.21)0.89 (0.77–1.03)1.14 (0.98–1.32)0.98 (0.81–1.18)
Age: ≥65 years1.35 (1.10–1.67)1.26 (1.03–1.56)1.00 (0.79–1.27)1.51 (1.13–2.02)1.07 (0.8668–1.32)1.48 (1.19–1.83)0.85 (0.69–1.06)1.39 (1.12–1.72)1.10 (0.84–1.44)
General health: fair0.97 (0.84–1.13)0.91 (0.78–1.05)0.95 (0.80–1.12)0.82 (0.67–1.01)1.04 (0.89–1.21)0.90 (0.78–1.05)0.96 (0.82–1.11)1.06(0.84–1.35)1.20 (0.99–1.47)
General health: bad/very bad0.80 (0.64–1.00)0.60 (0.48–0.75)0.78 (0.59–1.02)0.74 (0.53–1.04)1.22 (0.97–1.53)0.77 (0.61–0.98)0.84 (0.66–1.07)1.08 (0.86–1.37)1.18 (0.87–1.59)
Working status: employed0.98 (0.85–1.12)0.77 (0.68–0.88)0.99 (0.85–1.16)1.07 (0.88–1.29)0.85 (0.74–0.98)0.85 (0.75–0.98)1.01 (0.88–1.16)0.93 (0.81–1.07)0.85 (0.71–1.02)
Country: Wales0.98 (0.74–1.31)0.81 (0.61–1.07)0.98 (0.71–1.35)1.30 (0.88–1.90)0.96 (0.71–1.28)1.13 (0.85–1.51)0.97 (0.72–1.30)0.80 (0.59–1.10)0.90 (0.60–1.35)
Country: Scotland0.97 (0.79–1.20)1.20 (0.97–1.48)1.04 (0.82–1.32)1.16 (0.87–1.55)0.90 (0.72–1.13)1.10 (0.89–1.37)0.84 (0.67–1.05)1.01 (0.80–1.26)1.25 (0.95–1.64)
Scenario-specific explanatory variables
Children in household: yes-----1.63 (1.41–1.89)---
Reliance on carers: yes---1.30 (1.03–1.64)-----
Social grade: C2DE----1.36 (1.19–1.55)----
Model X2, p-value0.0575*<0.00010.7908*0.0161<0.0001<0.00010.1990*0.01520.0683
Deviance goodness of fit X2, p-value0.08850.20430.10820.12200.15160.05610.27650.0105*0.5012

4 DISCUSSION

Our study suggests, all else being equal, that severity of disease, diseases for which no other available treatments exist (representing unmet needs) and medicines that reduce reliance on informal carers (representing wider societal benefits) are supported by society as valid NHS resource prioritisation criteria. In the absence of other differences in patient or disease characteristics, or treatment effectiveness or costs, there were no preferences for any of the other prioritisation criteria we explored.

Under health benefit trade-off conditions there was, in all cases, a statistically significant shift in preferences towards the populations that gained a considerable improvement in health and away from the population that gained a little improvement in health, as we hypothesised. However, counter to our hypothesis, under cost (patient number) trade-off conditions, there was, with the exception of severity of disease, a statistically significant shift in preferences for all criteria towards the populations that were more costly to treat. Unless a preference was apparent under the scenario of ‘all else being equal’, the most plausible interpretation of these cost trade-off findings is that respondents are expressing a general preference for fairness in access to treatment based on need, irrespective of ability to benefit or cost, rather than a preference for the criterion in question per se. This is evident in the distributions of actual budget allocations made by respondents (Web Appendix III). For those criteria for which a societal preference was found under conditions of ‘all else being equal’, the distribution of budget allocations when costs were doubled were similar to when costs were equal between the competing populations, consistent with a clear preference for these criteria. However, for the remaining criteria, the actual budget allocations suggest that the cost difference causes a shift in budget allocations that peaks where around 70% of the budget is allocated to the most costly population. This is the nearest point to an equal split in patient numbers that our budget allocation scale would permit.

Our study therefore demonstrates that preferences are sensitive to the health gains that may be realised and the number of patients who may be treated, in contrast to our primary hypothesis that was grounded in the utilitarian view of population health (QALY) maximisation. Equity-efficiency trade-offs are being made by respondents, which may be driven by genuine specific social (or private) value judgements and/or more general, egalitarian principles of fairness.

4.1 Policy implications

4.1.1 Value-based pricing

Our study provides compelling evidence of societal support for all four proposed VBP criteria for rewarding new medicines with higher prices. Given that the government's consultation on the VBP system generated only eight (4%) responses from individual members of the public [Department of Health, 2011], the findings of support for these VBP criteria in our study, based on a sample of over 4000 members of the public, is reassuring.

Our study was not intended to provide specific weights or to define the levels at which the proposed VBP criteria should be applicable. These are among many other operational issues that remain to be resolved before adoption of VBP in the UK in 2014 [Hughes, 2011; Webb, 2011]. However, our study does confirm the importance of, and societal support for, the proposed criteria for which such weights and levels may need to be determined. On the basis of our findings, medicines that work in new ways are only valued above others when they produce a substantial health gain, and society is, at least in principle, supportive of the NHS paying more for innovative medicines that deliver substantial additional health benefits. But with median QALY gains observed in past AWMSG [Linley and Hughes, 2012] and SMC [Walker et al., 2009] submissions being of the order of only 0.1 QALYs, and evidence from published cost utility analyses suggesting incremental benefits of new interventions are decreasing over time [Greenberg et al., 2010], it remains to be seen how many new medicines that are declared to be innovative by manufacturers will be rewarded as such under VBP.

4.1.2 NICE criteria

NICE suggests that the six criteria it has identified to date as warranting special consideration in resource allocation decisions reflect societal preferences, as they are based predominantly on the views of its Citizens Council [Rawlins et al., 2010]. Although disease severity and significant innovation were supported in our study, we observed no compelling evidence for the three other prioritisation criteria we explored (disadvantaged populations, children and patients at the end-of-life). In reference to its end-of-life criterion, NICE states that the public generally places special value on treatments that prolong life at the end-of-life as long as that extension is of reasonable quality [Rawlins et al., 2010]. However, the Citizens Council report Departing from the threshold, 2008, indicates that, whilst 24 out of 29 (83%) council members favoured special consideration for treatments that are life saving, only 10 (35%) supported this view for treatments that are life extending [National Institute for Health and Clinical Excellence, 2008c]. This is consistent with our study, in which only 34% of respondents favoured prioritising patients with a reduced life expectancy in the absence of any other differences. Calls for a more systematic and transparent appraisal of medicines [Department of Health, 2010d], therefore, seem justified.

4.1.3 Cancer drugs fund

On the basis of the anticipated annual costs (8000 QALYs) and returns (4000 QALYs) of the CDF [Department of Health, 2010b], the government assumes society values health benefits to cancer patients at least twice as highly, all else being equal, than benefits to patients suffering other conditions. There was no robust empirical evidence in support of this assumption when the CDF was introduced and our study now provides empirical evidence to refute this assumption.

Several reports and studies have highlighted a lower uptake of new cancer medicines [Mason et al., 2010; Richards, 2008] and evidence of lower survival rates in the UK compared with other countries [Coleman et al., 2011]. However, the government's consultation document on the CDF points to delayed diagnosis as the main cause of poorer outcomes for cancer patients [Department of Health, 2010e]. A recent King's fund report agrees, adding that it is more important to improve access to surgery and radiotherapy, and noting that accessibility of cancer drugs is unlikely to have a significant overall impact [Foot and Harrison, 2011]. The consequence of a ring-fenced CDF is that funds are diverted away from services that overall may serve the wider population better, including many patients with cancer [Hughes and Duerden, 2011]. Our study, therefore, challenges the rationale for the CDF, which was introduced in England at a time when austerity measures were being actively imposed on other areas of the NHS [Department of Health, 2009].

4.1.4 Orphan drugs

In addition to being intended for the treatment of rare diseases, medicines that meet requirements for orphan drug designation [Fontain and Hemila, 2000] should address an unmet need, be indicated for life threatening or seriously debilitating (i.e. severe) conditions and may also meet a definition of significant innovation, all of which are supported as valid priority-setting criteria in our study. However, new medicines for the treatment of common, serious diseases may also address unmet needs and represent significant innovations [McCabe et al., 2005], so the issue of whether orphan drugs warrant special funding status would seem to rest on the value attached to rarity of disease.

Our study supports evidence from Norway [Desser et al., 2010] in finding no evidence of a societal preference for treating rare diseases over common diseases. In the absence of other compelling reasons for awarding special funding status to rarity per se [McCabe et al., 2005], the premise of specific orphan and ultra-orphan drug policies [All Wales Medicines Strategy Group, 2011; NHS Specialised Services, 2011; Scottish Medicines Consortium, 2010] appears open to question.

4.2 Strengths and limitations

Our study is the first comprehensive empirical analysis of societal views on issues that are central to UK policies on medicines reimbursement. Given the current austerity measures imposed on the NHS, and the imminent reforms of appraisal and reimbursement systems in the UK that have the potential to impact pharmaceutical pricing in other countries, our study is timely and informative for policy makers and national decision makers in the UK and further afield.

Our study has a number of methodological strengths. Our sample is broadly representative of the population of Great Britain and uses a larger sample than the UK population survey used to derive the EQ-5D tariff underpinning QALY calculations used by NICE (n = 3395) [Kind et al., 1999]. The format adopted for eliciting preferences has potential advantages over a simple binary choice question, by making participants more cognisant of the consequences of their decisions under trade-off conditions. As we explored trade-offs in both health gains and costs, a more complete picture of respondent trade-off behaviours is provided than using either health gains or costs alone.

There are some important caveats, however. Non-completion rates and details of non-responders were unavailable for analysis. This precluded any assessment of potential bias [Johnson and Wislar, 2012]. As in all choice-based experiments, participants were asked to make choices between hypothetical scenarios that inevitably involve simplification of complex decision problems. To avoid imposing our own interpretation of the prioritisation criteria, we constructed scenarios to reflect as closely as possible their definitions in guiding policies and criteria, but these may also simplify decision problems. For example, we defined unmet need in the context of no alternative treatments available, as per the VBP consultation document [Department of Health, 2010d]; however, the NHS would always provide some level of care, even if that is limited to symptomatic and palliative care. We also cannot be certain that respondents' preferences are not confounded by their own interpretations of the hypothetical scenarios.

It is possible that a central tendency bias exists in responses; however, we are reassured by the fact that respondent preferences, and shifts in preferences under trade-off conditions, are clearly differentiated across the nine criteria explored using the same elicitation method. Our study excludes preferences for situations where multiple criteria may coincide; however, none of the resource allocation criteria identified by NICE and proposed in the VBP consultation document are presented as being contingent on each other, and we still capture all criteria relevant to orphan drug designation, albeit separately.

As the UK NHS is a social insurance scheme that provides health care that is largely free to all at the point of access, the extent to which UK societal preferences would reflect the values of populations in other jurisdictions where other health care systems operate is unclear. We framed questions to encourage expressions of societal rather than private views, although our analyses suggest that respondent preferences may still be influenced by their personal circumstances. Some of these relationships have obvious, plausible explanations but others, such as those observed for respondents rating themselves as in bad/very bad health, are difficult to explain. Adaptation, comparison processes and cognitive dissonance [Stiggelbout and de Vogel-Voogt, 2008] may confound responses of those experiencing ill health and, as such, these findings should be interpreted with caution.

As may be anticipated due to the web-based research methods, our sample was slightly under representative of people aged over 65 years. Our sample was also possibly under representative of people in employment. Given that those over 65 years of age account for a greater proportion of health and social care spending, and those in employment, may feel they contribute to the funding of the NHS via taxation to a greater extent than those not in employment, under-representation of these groups' views could be potentially important. However, the degree to which these groups are under-represented is small. Our conclusions on societal preferences are based on majority views that may not reflect the intensity or ethical desirability of views. In mitigation, our large sample of respondents permits a broad range of potential views to be expressed.

Lastly, some commentators may consider that responses elicited via self-administered internet-based questionnaires are unreliable; however, cold elicitation methods, such as ours, may provide more representative accounts of preferences than other methods such as face-to-face interviews or deliberative process [Dolan and Tsuchiya, 2007] that have the potential to distort respondent preferences due to interviewer or group pressure bias [McColl et al., 2001].

4.3 Conclusions

The UK NHS has legal and moral obligations to provide fair, comprehensive, needs-based care for all. In doing so, it must also provide the best value for tax payers' money and the most effective, sustainable use of finite resources [Department of Health, 2010a]. Our study indicates that the public supports trade-offs in equity and efficiency in the allocation of scarce health care resources, but several prioritisation criteria currently imposed upon the NHS by NICE and the government do not reflect societal preferences. This may lead to inappropriate resource allocation decisions with significant population health and economic consequences, the benefactors being pharmaceutical manufacturers who are rewarded with higher prices for their medicines than may be warranted by the benefits they deliver. VBP aims to address these issues, and the proposed criteria for rewarding medicines with higher prices under this system do appear to have societal support.

CONFLICT OF INTEREST DISCLOSURE

WGL (operating via newmedinfo Ltd at the time of writing) and DAH produce the economic components of the AWMSG assessment reports for, and on behalf of, the All Wales Therapeutics and Toxicology Centre that may have an interest in the submitted work. DAH is a past member of the NICE technology appraisal committee and past deputy member of the AWMSG, which may have an interest in the submitted work; their spouses, partners or children have no financial relationships that may be relevant to the submitted work; WGL and DAH have no non-financial interests that may be relevant to the submitted work.

5 ACKNOWLEDGEMENTS AND DISCLOSURES

5.1 Acknowledgements

The authors wish to thank all pilot respondents and all anonymous respondents to the final questionnaire, Prof Rhiannon Tudor Edwards for contributing to early discussions and for providing comments on the draft manuscript, Dr Catrin Tudur Smith for the useful discussions on statistical analysis, and the two anonymous referees for their constructive comments. The authors alone are responsible for the resulting paper.

5.2 Financial disclosure

Funding for the project was provided by a Bangor University PhD studentship awarded to WGL. The funder had no role in study design, data collection and analysis, decision to publish or preparation of the manuscript.

5.3 Author contributions

WGL and DAH conceived the study and designed the survey. WGL commissioned data collection, managed and analysed the data. WGL and DAH interpreted the results. WGL drafted the manuscript. WGL and DAH critically revised the manuscript for intellectual content. WGL and DAH approved the final version of the paper. DAH had full access to all data, had final responsibility for the decision to submit for publication and is a guarantor.

5.4 Competing interests

WGL and DAH have support from Bangor University for the submitted work; WGL (operating via newmedinfo Ltd at the time of writing) and DAH produce the economic components of the AWMSG assessment reports for, and on behalf of, the All Wales Therapeutics and Toxicology Centre, which may have an interest in the submitted work. DAH is past member of the NICE technology appraisal committee and past deputy member of the AWMSG, which may have an interest in the submitted work; their spouses, partners, or children have no financial relationships that may be relevant to the submitted work; WGL and DAH have no non-financial interests that may be relevant to the submitted work.

5.5 Ethical approval

We obtained approval for the study from Bangor University research ethics committee (ref: 2011–3641).