Estimation of Utilities in Attention-Deficit Hyperactivity Disorder for Economic Evaluations

Estimation of Utilities in Attention-DeficitHyperactivity Disorder for EconomicEvaluationsAndrew Lloyd,1 Paul Hodgkins,2 Rahul Sasane,2 Ron Akehurst,3 Edmund J.S. Sonuga-Barke,4

Patrick Fitzgerald,3 Annabel Nixon,1 Haim Erder2 and John Brazier3

1 Oxford Outcomes Ltd, Oxford, UK

2 Shire Pharmaceuticals, Wayne, PA, USA

3 Health Economics and Decision Science, School of Health and Related Research, University of Sheffield,

Sheffield, UK

4 School of Psychology, Southampton University, Southampton, UK

Abstract Background: Attempts to estimate the cost effectiveness of attention-deficit

hyperactivity disorder (ADHD) treatments in the past have relied on classi-

fying ADHD patients as responders or non-responders to treatment. Res-

ponder status has been associated with a small gain in health-related quality

of life (HR-QOL) [or utility, as measured using the generic QOLmeasure EQ-

5D] of 0.06 (on a scale from 0 being dead to 1.0 being full health).

Objectives: The goal of the present study was to develop and validate several

ADHD-related health states, and to estimate utility values measured amongst

the general public for those states and to re-estimate utility values associated

with responder status.

Methods: Detailed qualitative interview data were collected from 20 young

ADHD patients to characterize their HR-QOL. In addition, item-by-item

clinical and HR-QOL data from a clinical trial were used to define and describe

four health states (normal; borderline to mildly ill; moderately to markedly ill;

and severely ill). ADHD experts assessed the content validity of the descrip-

tions. The states were rated by 100 members of the UK general public using the

time trade-off (TTO) interview and visual analog scale. Statistical mapping was

also undertaken to estimate Clinical Global Impression-Improvement (CGI-I)

utilities (i.e. response status) from Clinical Global Impression-Severity (CGI-S)

defined states. The mapping work estimated changes in utilities from study

baseline to last visit for patients with a CGI-I score of £2 or £3.Results: The validity of the four health states developed in this study was

supported by in-depth interviews with ADHD experts and patients, and

clinical trial data. TTO-derived utilities for the four health states ranged from

0.839 (CGI-S state ‘normal’) to 0.444 (CGI-S state ‘severely ill’). From the

mapping work, the change in utility for treatment responders was 0.19 for

patients with a CGI-I score of £2 and 0.15 for patients with a CGI-I score of £3.

ORIGINAL RESEARCH ARTICLEPatient 2011; 4 (4): 247-257

1178-1653/11/0004-0247/$49.95/0

ª 2011 Adis Data Information BV. All rights reserved.

Conclusions: The present study provides utilities for different severity levels of

ADHD estimated in a TTO study. This approach provides a more granular

assessment of the impact of ADHD on HR-QOL than binary approaches em-

ployed in previous economic analyses. Change in utility for responders and non-

responders at different levels of CGI-I was estimated, and thus these utilities

may be used to compare health gains of different ADHD interventions.

Key points for decision makers

� Previous assessments of the cost effectiveness of treatments for ADHD have often reliedupon a binary classification of patients as responders or non-responders

� The current work was designed to provide utility estimates reflecting different levels of severityof the patient’s symptoms

� The estimated utility values may offer improved sensitivity and specificity for economic eval-uations

� Utilities for responder and non-responder health states were also estimated to providecompatibility with previous economic evaluations

Background

Attention-deficit hyperactivity disorder (ADHD)is the most common childhood onset neuro-behavioral disorder. It has been estimated that5% of school-aged children and adolescents wouldmeet the Diagnostic and Statistical Manual ofMental Disorders-IV (DSM-IV)[1] criteria forADHD, which is equivalent to 366 000 childrenand adolescents in England and Wales.[2] ADHD isa chronic, debilitating disorder defined by devel-opmentally inappropriate hyperactivity, impulsivity,and inattention. ADHD commonly co-occurs withother childhood disorders, such as oppositionaldefiant disorder, conduct disorder, or learning dis-abilities.[3-5] Children with ADHD often have lowself-esteem, develop emotional and social problems,and frequently underachieve at school.[2] Moregenerally, children with ADHD often struggle tomanage the demands of school.[6] Reading ageand scholastic achievement is low and absenteeismand failure to graduate from school are high.[7]

ADHD is also associated with anxiety,[8,9] de-pression, aggression, and sleep problems.[10,11]

Impairment with social functioning, a central fea-ture of ADHD,[1] can include two characteristic

types of behavior: negative/aggressive nature of thechild’s interactions, and the child’s hyperactivity orimpulsivity.[12]

Health-related quality of life (HR-QOL) iswidely accepted to be a broad concept that com-prises an individual’s subjective perception of theimpact of a disease, and its treatment, on daily life,physical, psychological, and social functioning, andwell-being.[13] Some quantitative investigations intothe HR-QOL of children and adolescents withADHD have been reported.[11,14-19] HR-QOL issubstantially reduced in ADHD patients comparedwith other children their own age;[15] studies havedemonstrated the benefit of pharmacotherapy forADHD in terms of HR-QOL.[18-21]

HR-QOL is an important endpoint for as-sessing efficacy and cost effectiveness of ADHDtreatments. Pivotal trials for regulatory approvalof drugs may focus on symptom reduction asthe primary outcome, but economic evaluationsembracing the ‘value’ of the treatment for re-imbursement also considerHR-QOLand functionalimpairment. For instance, the National Institutefor Health and Clinical Excellence (NICE) in theUK prefer outcomes to be measured in terms ofquality-adjusted life-years (QALYs). A QALY

248 Lloyd et al.

ª 2011 Adis Data Information BV. All rights reserved. Patient 2011; 4 (4)

represents the product of time in a state or survivaland HR-QOL. However, to estimate a QALY,NICE state that HR-QOL must be expressed on asingle index that reflects the value of a health stateand has cardinal measurement properties.[22] Theeconomic concept of utility can be used to reflectthe strength of preference for a state of health orHR-QOL. Value can be expressed as a utility usingthe time trade-off (TTO) method where people in-dicate the value of a state in terms of years of lifethey are willing to give up to avoid it.

NICE have previously reviewed treatments forADHD, and in their economic evaluations, peoplewith ADHDwere either classified as responders ornon-responders based upon the Clinical GlobalImpression-Improvement (CGI-I), with a score of1 or 2 only classified as a response.[2] However, theCGI-I is a relative measure of change with thepotential outcome that some individuals could bemuch improved but still remain highly impaired.Previous work estimated utility values for the twohealth states as 0.837 for responders and 0.773 fornon-responders based upon parent’s proxy ratingsof their child’s HR-QOL.[23] NICE concluded intheir technology assessment that further research isrequired to better capture the utility values asso-ciated with ADHD.[2] Subsequently, NICE havestated that they prefer utilities derived from a ge-neric HR-QOL measure, the EQ-5D (previouslyknown as the EuroQol), but recognize that it is notdesigned for use with children.[22] One alternativeto the use of the EQ-5D is the use of disease-specific utility methods.

The goal of the present study was to generatenew evidence regarding utility values in childhoodADHD. The study was designed to define andvalidate different ADHD health states based onvignette descriptions and to estimate utilities usinga combination of different methods, with the in-tention of better quantifying preferences for healthgains associated with treatment of ADHD.

Methods

Health State Development

Health states were developed through com-bining two methodologies: (i) in-depth interviews

with young people with ADHD and the parentsof young people with ADHD; and (ii) analysis ofbaseline clinical trial data from a trial of amethylphenidate-based therapy. A general re-view of the literature was undertaken in order toguide the development of discussion guides forthe in-depth interviews.

Qualitative Interviews

Participants were recruited from responsesto newspaper advertisements and from localADHD support groups. Informed consent wasobtained from all participants prior to completinginterviews. Subjects were 20 children with ADHD(aged 11–16 years), who had received (within thepast 1 year) prescription medication or a behav-ioral therapy-based treatment for ADHD for atleast 1 month. Parents or primary carers (n= 17,nine completed as parent-child dyad) of childrenwith ADHDwere also interviewed to capture dataon the impact of ADHD on their child (parent/primary carer were required to have been in thatrole for ‡5 years). Interview guides included ques-tions regarding how ADHD affected HR-QOL.Semi-structured interviews were undertaken us-ing open-ended questions and appropriate probesby experienced interviewers. Data regarding demo-graphics, education, ADHD symptoms, medi-cation, and therapy-based treatments were alsogathered. Parents/carers completed the Conners’ParentRating Scale-Revised short form (CPRS-R).The adolescent participants completed the Con-ners’-Wells’ Adolescent Self-Report Scale shortform. A thematic qualitative analysis was per-formed on the transcribed interview data toidentify emergent themes and concepts.[24]

Clinical Trial Analysis

Baseline clinical trial data from a trial of amethylphenidate-based therapy[25] were analyzedto support the development of the health states.This trial included different outcomes assessmentsincluding the Clinical Global Impression-Severity(CGI-S),[26] the ADHD Rating Scale (ADHD-RS),[27] and the ADHD Impact Module-Child(AIM-C). The CGI-S is a 7-point clinical globalimpression scale where the physician rates theseverity of the patient’s illness (1 being normal,

Estimating Utilities in ADHD 249


not at all ill; 2 being borderline ill; 3 being mildlyill; 4 being moderately ill; 5 being markedly ill;6 being severely ill; 7 being extremely ill). TheADHD-RS is a widely used scale to diagnosepatients with ADHD and to assess treatmentresponse. The AIM-C is a survey measure specif-ically designed to capture the impact of ADHDon the child and their QOL.

Trial participants were grouped according totheir baseline CGI-S level (e.g. ‘normal’, ‘bor-derline ill,’ etc.). The frequency of responses toindividual questions on the ADHD-RS[27] andthe AIM-C[20] was analyzed. For example, on theAIM-C, this is the frequency of endorsing theresponse choice ‘a lot,’ ‘quite a bit,’ ‘some,’ etc.The most frequent response choice statement wasincluded as text in the health states (e.g. ‘Some-times you find it difficult to sit still.’).

Health States

Short descriptions of potential health stateswere developed based on the clinical trial anal-ysis. Final health states were derived after con-sideration of qualitative data from the in-depthinterviews and discussion amongst members ofthe study team, including an ADHD expert. Thefinalized health states were used in a cognitivedebrief interview conducted with members of thegeneral public to ensure the vignettes adequatelydifferentiated between the health states, andwere clear and easy to understand. The healthstates were then included in the valuationexercise.

Health State Valuation

A convenience sample of members of the public(n= 100) were recruited from four geographicallydistinct locations across the UK to rate the healthstates. After providing informed consent, eachparticipant completed a background questionnaireand the EQ-5D (a measure of HR-QOL). Partic-ipants then rated each health state using the100-point visual analog scale (VAS) and TTOtechniques.[28] Participants were not informedthat the states were designed to describe ADHDand the states were not labeled. The VAS was usedprimarily to familiarize people with the concept

of rating health states. In the TTO task, partic-ipants were asked to imagine that they had ahealth state as described on the card. They thenchose whether they preferred (i) to live in thehealth state for 10 years followed by death; (ii) tolive in 10 - x years in full health; or (iii) to indicatethat the two previous options were equally pref-erable. Full health was described as no problemswith mobility, self-care, usual activities, no painor discomfort, and no anxiety or depression (whichis essentially the EQ-5D state for full health). Theamount of time in full health was then variedsystematically (to avoid anchoring bias[29]), untilthe participant was indifferent between 10 yearsin the health state or 10 - x years in full health. Allparticipants were offered compensation (d25) fortheir time.

Mapping the Severity Scale to theImprovement Scale to Estimate TreatmentResponse

Decision analyticmodels forADHDcommonlyuse ameasure of outcome or change between statesas measured by the CGI-I scale rather than theCGI-S scale. Therefore, mapping work was under-taken so that utilities for the CGI-I scale could beestimated from a measure of CGI-S. In addition,the mapping work was designed to also allow us toreport the current results in terms of the responder/non-responder format used previously by NICEand other health technology assessors.

The mapping methods were designed to estab-lish a statistical relationship between the CGI-Iand the CGI-S. The utility values that had beendetermined for levels of CGI-S could then be usedto estimate utilities for CGI-I based on the map-ping function. The estimation of the mappingfunction was based on data from two clinicaltrials,[25,30] which were combined for this analysis.CGI-S values at follow-up were estimated using alinear regression of CGI-S scores (range 1–7) onADHD-RS at baseline for the combined trials.The resulting predicted CGI-S scores were tabu-lated by whether or not patients had responded totreatment where response to treatment was de-fined as having achieved either the top two (‘verymuch improved’ or ‘much improved’) or the top

250 Lloyd et al.


three scores (‘very much improved’, ‘much im-proved,’ or ‘minimally improved’) in the CGI-Iscale at the last visit. Mean TTO utility values forresponders and non-responders, plus correspond-ing standard errors, were calculated by a weightedaverage of the CGI-S score frequencies.

Results

Qualitative Interview Sample

The children with ADHD had the followingprofile: mean (SD) age was 11 (2.5) years (range8–16 years); 80% male; mean (SD) age of symp-tom onset was 3 (1.47) years (range 0.5–7 years);andmean (SD) age at diagnosis was 8 (3.08) years(range 4–15 years). By parent/carer report, thechildren had combined type ADHD (40% of sub-jects), predominately hyperactive-impulsive sub-type (25%) or predominantly inattentive sub-type(5%); or it was unknown (30%). CPRS-R scoresranged from 45 to 90 (mean–SD 70.75– 13.97)for the oppositional scale, and 59–90 for the cog-nitive (mean– SD 71.0– 8.56) and hyperactivity(mean– SD 80.7– 10.5) scales, indicating a clin-ically significant problem.[31] ADHD index scores(reflecting Conners-Wells’ Self-Report Scale andConners’ Parent Rating Scale-Revised) for partic-ipants ranged from 57 to 90 (mean– SD74.3– 9.25).Parents/carers also reported problems at schoolthat included the child being disciplined (80%),being required to repeat a year (15%), or beingexpelled (15%).

The qualitative analysis of the data identifiedmany of the common features of ADHD. Childrenwith ADHD liked physical education, swimmingand climbing, and physical activities, which insome cases improved mood and concentration.Children were often unable to sit still, fidgeted,and talked incessantly. They also could be clumsyand accident prone, including injuring themselves.Children had difficulty settling at night, and sometook up to 3 hours to fall asleep. Children hadpoor concentration and described themselves asbeing easily distracted and forgetful, as well ashaving poor listening, planning, and organiza-tional skills. In school, the children had difficultieswith written work andmade simple mistakes. They

found it difficult to organize themselves for schooland had difficulty completing their homework.

The children were generally happy, but theirmoods were very changeable. Parents/carers re-ported that the children were easily provokedto anger, and became annoyed and frustratedwith those around them. Parents/carers also re-ported that the children often did not have self-confidence or self-esteem. Children’s social skillswere not well developed, and so they had diffi-culties relating to their peer group. Many chil-dren felt that having few friends was difficult forthem, and several reported feeling lonely; theirhyperactive and impulsive behaviors led to thembeing frequently excluded from social clubs aswell as from school. Many of the children withADHD reported being teased or bullied by peers.

Clinical Trial Population

The trial was a single-arm, open-label trial. Theprofile of the sample (n= 171) was as follows:meanage 9.4 years (SD 1.9); 71% male; 79% WhiteCaucasian, 12% African American, 9% other eth-nicity; duration of ADHD= 3.6 years (SD 2.1);ADHD type: 77% combined, 21% inattentive, 2%hyperactive; mean ADHD-RS = 14.1 (SD 7.5);mean CPRS-R= 77.1 (SD 45.4) [see Arnold et al.[25]

for more details].

Health States

The clinical trial analysis process produced alot of statements for potential inclusion in eachhealth state. This information was also con-sidered alongside the qualitative data fromthe interviews and was discussed with expertsto produce the content of the final states. Forexample, one item in the ADHD-RS, ‘‘Has diffi-culty awaiting turn’’: parents of children catego-rized as ‘mildly ill’ most frequently responded‘‘Sometimes’’ to this statement. In the resultinghealth state, this was combined with informationfrom the qualitative interviews to produce ‘‘Youare sometimes impulsive and find it difficult towait your turn.’’ Not all items in the AIM-C andADHD-RS could be included in the health statesbecause the vignettes (short descriptions of eachhealth state) would be too long. Therefore, items



from the AIM-C and ADHD-RS that were add-itionally identified in the qualitative interviewdata were included in the health states, usingsimilar methods to Lloyd et al.[32]

Health states describing six levels of CGI-S(apart from the most severe level of CGI-Sbecause no data were available on this) weredeveloped (table I). The health states included adescription of the main dimensions of HR-QOL(such as physical, psychological, and socialfunctioning), as well as the specific ADHDsymptom and functioning burden that was iden-tified. The health states did not make specificreference to a pediatric condition. The healthstates were designed to be a fair reflection of thedata that were available.

Cognitive Debriefing

Members of the study team, including an ex-pert in ADHD (ESB), reviewed the draft healthstates to evaluate their content. There was ageneral consensus that the differences betweensome of the states were minor and it was difficultto distinguish between them. Therefore, it wasdecided to simplify the six states into just fourstates without losing important information. Twonew states were developed (‘borderline to mildlyill’ and ‘moderately to markedly ill’) by mergingthe existing descriptions. Sixmembers of the generalpublic completed a cognitive debrief interview todetermine if they understood the content of thestates and could recognize differences betweenthem. All of the participants indicated that fourhealth states (as opposed to six) were easier todifferentiate. The content of the states was well

understood, and no further clarifications orchanges were required following the cognitivedebrief interviews (see the Supplemental DigitalContent for health state descriptions, http://links.adisonline.com/PBZ/A32).

Health State Valuation Study Sample

On most sociodemographic and backgroundvariables, the study sample was well matched tothe UK population (tables II and III). The studysample was over-represented in terms of peoplewho described themselves as Asian, and had ahigher than expected proportion of students anda high proportion of people with university-level

Table I. The initial six health states that were drafted based on pa-

tient/carer interviews and clinical trial analysis, and the final four that

were included in the time trade-off valuation study with the general

public

Initial CGI-S states Final study states

Normal Normal

Borderline ill Borderline to mildly ill

Mildly ill

Moderately ill Moderately to markedly ill

Markedly ill

Severely ill Severely ill

CGI-S =Clinical Global Impression-Severity.

Table II. Background characteristics of the general public sample

included in the time trade-off valuation study compared with UK

census/Office of National Statistics (ONS) dataa

Characteristic Participants

(n =100)UK census data

2003/ONS data

Age (y) [mean (SD)] 38.5 (15.01) 38.2

Sex (% female) 53.0 51.0

Ethnic group

Asian or Asian British 13.0 4.0

Black or British Black 1.0 2.0

Chinese, mixed, or other 0.0 2.0

White 86.0 92.0

Employment status

full time 57.0 73.8b

part time 9.0 NA

student 12.0 2.0

seeking work 8.0 6.8b

retired 8.0 10.0

stay at home 3.0 5.0

other 3.0 NA

Education – leaving age

left school with no

qualifications

7.0 NA

left school aged 16 y with

qualifications

12.0 NA

stayed on at school/college to

gain further qualifications

15.0 NA

vocational qualifications 27.0 NA

university level 38.0 NA

prefer not to answer 1.0 NA

a All values are presented as % unless otherwise indicated.

b Data from April 2009.

NA =not available.

252 Lloyd et al.


education. Our sample also reported slightlybetter HR-QOL (as measured by EQ-5D) thanseen in data from a previous UK national sur-vey.[33] This was particularly evident in terms ofmobility (table III).

Health State Valuation

Table IV shows themean VAS values and TTOutilities for each health state. The VAS and TTOresults show how severely people rated the declinein HR-QOL associated with worsening ADHD.Participants rated the ADHD ‘normal’ state onTTO as 0.839. The first two states (‘normal’ and‘borderline to mildly ill’) were rated quite sim-ilarly, while the remaining two states were ratedas much more severe using both VAS and TTOmethods.

Estimated Treatment Response fromMapping

From the mapping work, the estimated re-gression equation was CGI-S = 0.1138 ·ADHD-RS with the result rounded to the nearest integer(range 1–7). The goodness-of-fit index was sat-

isfactory (R2 = 0.934). Resulting mean utilityvalues for responders and non-responders, pluscorresponding standard errors, were calculatedusing a weighted average of the CGI-S score fre-quencies. Table V presents the resultant TTOutility values at baseline and follow-up for res-ponders and non-responders (defined both ways)that could be used to populate ADHD cost-effectiveness models.

Discussion

The present study reports three different setsof utility values describing ADHD-related healthstates. The study was designed to provide betterquality utility estimates than those that are cur-rently available. In addition, the study was alsodesigned to produce values that would be suitedto different approaches in economic evaluations.

The first set of values was generated from theTTO study. Health state vignettes were devel-oped, based upon clinical trial data and in-depthpatient/parent interviews, to describe differentseverities of ADHD defined in terms of levels of

Table III. General public participants’ ratings of their current health from the EQ-5D compared with data from UK national surveys reported by

Kind et al.[33]

Dimension Study sample (%) [n =100] UK norms (%) [n =3395]

moderate problem extreme problem moderate problem extreme problem

Mobility 2.0 0.0 18.3 0.1

Self-care 2.0 0.0 4.1 0.1

Usual activity 8.0 0.0 14.2 2.1

Pain/discomfort 22.0 0.0 29.2 3.8

Anxiety/depression 22.0 0.0 19.1 1.8

Table IV. Descriptive statistics for the visual analog scale (VAS) values and time trade-off (TTO) utility values for the four health states

(n= 100)

Health states VASa TTO

mean (SD) 95% CI mean (SD) 95% CI

Normal 73.3 (15.4) 70.3, 76.4 0.839 (0.203) 0.799, 0.880

Borderline to mildly ill 64.4 (15.8) 61.2, 67.5 0.787b (0.217) 0.744, 0.830

Moderately to markedly ill 45.1 (15.8) 41.9, 48.2 0.578c (0.275) 0.523, 0.633

Severely ill 37.6 (16.6) 34.3, 40.9 0.444d (0.230) 0.385, 0.504

a Range of the VAS =0–100.

b Significantly lower than ‘normal’ (p <0.01).

c Significantly lower than ‘borderline to mildly ill’ (p< 0.01).

d Significantly lower than ‘moderately to markedly ill’ (p< 0.01).



CGI-S. The multiple sources of information usedto develop the vignettes should support theircontent validity and represent an improvementover previous vignette studies, which often relyon expert opinion only. The TTO data produceda substantial range of scores from 0.839 to 0.444.The values for the states ‘normal’ (0.839) and‘borderline to mildly ill’ (0.787) were very close tothe values reported by Coghill et al.[23] and usedby NICE for ‘responder’ (0.837) and ‘non-responder’ (0.773). This provides some additionalexternal validation of utilities generated in thisstudy.

The values from the TTO study have severaladvantages, but they also have disadvantages. Intheir favor, the vignettes were based on differentdata sources, which were used together to de-velop the descriptions. The qualitative interviewsrevealed that the general public participants wereable to imagine themselves in the state described,and the TTOmethod was able to capture people’svaluations or preferences for these states. How-ever, there are some limitations with these data.Although the general public were able to under-stand and rate the health states developed, anincremental step in validation would have been toreview the health states with patients. Not allADHD studies use CGI-S as an outcome meas-ure but instead use the global impression of im-provement: the CGI-I. CGI-I measures relativechange and so cannot be used as a way of definingan absolute state. This means that the resultsfrom the TTO study reported here cannot alwaysbe applied in an economic evaluation, becauseADHD therapies are not always assessed in termsof changes in a CGI-S score. One other limitationis that the vignette approach is not a recommen-ded method for capturing utilities according to

NICE.[22] NICE are explicit that they prefer util-ities to be derived from a generic HR-QOL meas-ure such as the EQ-5D, which is completed by thepatient. However, NICE also state that the EQ-5Dmay not be appropriate for use in children.[22]

The present study also describes two more setsof utility values, which were derived from stat-istical mapping work. This work was undertakento estimate values for patients defined in terms ofCGI-I fromCGI-S scores and so allows the utilityscore from the TTO study to be more widelyused. The estimated utilities from the mappingwork could be used in economic evaluationswhere there is a need to compare new treatmentsagainst trial data from current therapies in termsof the CGI-I. In line with the approach used byNICE previously,[2] using the current analyses,values for responders and non-responders wereestimated at baseline and follow-up. The map-ping results have been applied using two differentcriteria to define responders. In one analysis,‘responder’ was defined in terms of levels 1 and 2on CGI-I (‘very much improved’ and ‘much im-proved,’ respectively) and, in the second analysis,the responder definition also included people atlevel 3 (‘minimally improved’) on the CGI-I. Thesecond analysis reduced the overall gain in utilityfrom baseline to study end, as would be expected.

This approach to estimating utilities based onmapping also has advantages and limitations.First, the mapping work allowed us to addresssome of the limitations of the TTO study. Theseutility data can be used to compare novel treat-ments against the current standard of care (interms of CGI-I) when CGI-S has been used tomeasure treatment changes. Furthermore, dif-ferent responder definitions that were definedallow a degree of flexibility for health economists

Table V. Mean (standard error) time trade-off (TTO) values by time-point for two response threshold scores by Clinical Global Impression for

Improvement (CGI-I) score derived from statistical mapping

Threshold: Final CGI-I score £2a Final CGI-I score £3b

Time-point: baseline last visit baseline last visit

Outcome: responder non-responder responder non-responder responder non-responder responder non-responder

TTO 0.63 (0.22) 0.65 (0.21) 0.82 (0.19) 0.70 (0.20) 0.64 (0.22) 0.67 (0.21) 0.79 (0.19) 0.67 (0.20)

a Response was a CGI-score of £2, i.e. CGI-I score of 1 or 2 (‘very much improved’ or ‘much improved,’ respectively).

b Response was a CGI-I score of £3, i.e. a CGI-I score of 1, 2, or 3 (‘very much improved’, ‘much improved,’ or ‘minimally improved,’

respectively).

254 Lloyd et al.


to use in modeling treatment benefits. This ap-proach has backward compatibility with previouseconomic evaluations and can be used for futureassessments. However, there are limitations. Theestimates are based on a mapping function and soare influenced by the robustness of that mappingfunction. More fundamentally, however, so as toretain the approach used byNICE previously, themethod classifies patients as either responders ornon-responders. One aim of the present researchwas to improve the sophistication and sensitivityof utility estimates in childhood ADHD, but theconstraints of existing analyses meant that theresults were presented in binary form. However,in principle, the approach could be used to alsogo beyond this simple dichotomy. Finally, it isalso worth noting that the mapping approachproduced a much larger change in utility frombaseline to end of study for responders than wasassumed by NICE.[2] Coghill et al.[23] reportedobserved data (albeit with limitations), while themapping approach produces estimated values, soit is possible that the mapping approach has ex-aggerated the differences in utility. Both ap-proaches have limitations, and it may not bepossible to determine which approach is mostaccurate until new prospective data are available.Sensitivity analyses in economic evaluations areencouraged to test the robustness of the results.

Despite these limitations, the study presentsimportant new information regarding the impactof ADHD on HR-QOL. The qualitative inter-views present a detailed impression of how sig-nificantly ADHD affects children. The observedutility values from the TTO study indicated thatthe general public considered these states to rep-resent a genuine impairment in HR-QOL. Theexisting literature in this area is limited. Coghillet al.[23] reported proxy-rated EQ-5D values,which were recognized as having limitations.[3]

Matza et al.[34] reported utility values from avignette study where 43 parents of children withADHD rated different health states. Utility val-ues ranged from 0.90 (severe untreated ADHD)to 0.98 (effective and tolerable non-stimulanttreatment), which were higher than estimatesfrom other studies (e.g. Coghill et al.[23]). Thevalues may be high because it was parents and not

the general public who rated the health states.Most decision makers such as NICE prefer pre-ference weights to come from the general public.Studies where patients rate their own health oftenproduce values much higher than the generalpublic rating the same states, probably becausepeople learn to adjust and adapt/cope with theirhealth status.[35] This effect may have also oc-curred in the Matza et al.[34] study to an extent.

Conclusions

To summarize, we developed and validatedhealth states, defined by the CGI-S, based upontrial data and qualitative information from chil-dren with ADHD. This meant that descriptionsreflected actual patient experience. Participantsin the valuation work could understand and ratethe detailed information in the health states. Weemployed mapping techniques so that utilities forhealth states defined by CGI-I could be esti-mated. This is an important dimension of thestudy that allows these values to be used tocompare new treatments against existing thera-pies. The resulting utilities are considered to haveface validity, and may offer improved sensitivityand specificity for economic evaluations.

Acknowledgments

The study was funded by Shire Pharmaceuticals, whichdevelops and markets treatment for ADHD. PH and HE areemployees and stockholders of Shire Pharmaceuticals. RS wasan employee of Shire Pharmaceuticals when the research wasundertaken. All three contributed to the conduct of the study,including interpretation of the data, and preparation, review,and approval of the manuscript. AL and AN are employees ofOxford Outcomes, and this company was paid a fixed fee toundertake the study. PF, JB, and RA are all employees of theUniversity of Sheffield who were also paid a fixed fee for theircontribution to the study. ESB has received consultancy pay-ments and honoraria from Shire Pharmaceuticals.

AL directed the study, developed the methodology, andtook the lead in writing the manuscript. PH conceived theneed for the study, contributed to the study throughout, andco-wrote the manuscript. RS contributed to the study meth-odology throughout and contributed to the manuscript. RAprovided a considerable degree of conceptual input at the startof the study and helped to interpret the study findings. ESBprovided expert insight into how HR-QOL is affected inADHD. He also co-wrote large parts of the paper. ANwas instrumental in the development of the qualitative study



design and planning for analysis. She also co-wrote parts ofthe manuscript. PF undertook the statistical mapping work.JB contributed to the conception of the study, helped developthe health states, supported the mapping work, and helped co-write parts of the paper.

References1. American Psychiatric Association. Diagnostic and statistical

manual of mental disorders. 4th ed. Washington, DC:APA, 1994

2. National Institute for Health and Clinical Excellence.Methylphenidate, atomoxetine and dexamfetamine for at-tention deficit hyperactivity disorder (ADHD) in childrenand adolescents. (TA98). London: NICE, 2006 Mar [on-line]. Available from URL: http://www.nice.org.uk/TA098[Accessed 2009 May 6]

3. Pliszka SR, Carlson C, Swanson JM. ADHDwith comorbiddisorders: clinical assessment andmanagement. NewYork:Guilford, 1999

4. Harpin VA. The effect of ADHD on the life of an individual,their family, and community from preschool to adult life.Arch Dis Child 2005; 90: 2-7

5. Biederman J, Newcom J, Sprich S. Comorbidity of atten-tion-deficit hyperactivity disorder with conduct, depres-sive, anxiety and other disorders. Am J Psychiatry 1991;148: 564-77

6. Barbaresi WJ, Katusic SK, Colligan RC, et al. Long-termschool outcomes for children with attention-deficit/hyperactivity disorder: a population-based perspective. J DevBehav Pediatr 2007; 28: 265-73

7. Kendall J, HattonD, Beckett A, et al. Children’s accounts ofattention-deficit/hyperactivity disorder. Adv Nurs Science2003; 26: 114-30

8. Schatz DB, Rostain AL. ADHD with comorbid anxiety: areview of the current literature. J Atten Disord 2006; 10:141-9

9. Levy F. Synaptic gating and ADHD: a biological theory ofcomorbidity of ADHD and anxiety. Neuropsychopharm-acology 2004; 29: 1589-96

10. Treuting JJ, Hinshaw SP. Depression and self-esteem inboys with attention-deficit/hyperactivity disorder: associa-tions with comorbid aggression and explanatory attri-bution mechanisms. J Abnorm Child Psychol 2001; 29:23-39

11. Ralston SJ, Lorenzo MJM. ADORE – Attention-Deficithyperactivity disorder Observational Research in Europe.Eur Child Adolesc Psychiatry 2004; 13 (1 Suppl.): i36-42

12. Nijmeijer JS, Minderaa RB, Buitelaar JK, et al. Attention-deficit/hyperactivity disorder and social dysfunctioning.Clin Psychol Rev 2008; 28: 692-708

13. Coghill D, Danckaerts M, Sonuga-Barke E, et al. Practi-tioner review: quality of life in child mental health – con-ceptual challenges and practical choices. J Child PsycholPsychiatry 2009; 50 (5): 544-61

14. Klassen AF, Miller A, Fine S. Agreement between parentand child report of quality of life in children with attention-deficit/hyperactivity disorder. Child Care Health Dev 2006;32: 397-406

15. Escobar R, Soutullo CA, Hervas A, et al. Worse quality oflife for children with newly diagnosed attention-deficit/

hyperactivity disorder, compared with asthmatic and healthychildren. Pediatrics 2005; 116 (3): e364-9

16. Topolski TD, Edwards DC, Patrick DL, et al. Quality of lifeof adolescent males with attention-deficit hyperactivitydisorder. J Atten Disord 2004; 7: 163-73

17. Buitelaar JK, Wilens TE, Zhang S, et al. Comparison ofsymptomatic versus functional changes in children andadolescents with ADHD during randomized, double-blindtreatment with psychostimulants, atomoxetine, or placebo.J Child Psychol Psychiatry 2009; 50: 335-42

18. Bukstein OG, Arnold LE, Landgraf JM, et al. Doesswitching from oral extended-release methylphenidate tothe methylphenidate transdermal system affect health-related quality-of-life and medication satisfaction for chil-dren with attention-deficit/hyperactivity disorder? ChildAdolesc Psychiatry Ment Health 2009; 3 (1): 39

19. Manos M, Frazier TW, Landgraf JM, et al. HRQL andmedication satisfaction in children with ADHD treatedwith the methylphenidate transdermal system. Curr MedRes Opin 2009; 25 (12): 3001-10

20. Landgraf JM, Rich M, Rappaport L. Measuring quality oflife in children with attention-deficit hyperactivity disorderand their families. ArchPediatrAdolescMed 2002; 156: 384-91

21. Meaux JB, Hester C, Smith B, et al. Stimulant medications:a trade-off? The lived experience of adolescents withADHD. J Spec Pediatr Nurs 2006; 11: 214-26

22. National Institute for Health and Clinical Excellence. Guideto the methods of technology appraisal. London: NICE,2008 June [online]. Available from URL: http://www.nice.org.uk/media/B52/A7/TAMethodsGuideUpdatedJune2008.pdf [Accessed 2009 May 6]

23. Coghill D, Spender Q, Barton J, et al. Measuring quality oflife in children with attention-deficit-hyperactivity disorderin the United Kingdom. 16th World Congress of the In-ternational Association of Child and Adolescent Psychiatryand Allied Professions (IACAPAP); 2004 Aug 23-26; Berlin

24. Joffe H, Yardley L. Content and thematic analysis. In:Marks D, Yardley L, editors. Research methods for clinicaland health psychology. London: Sage, 2004: 56-68

25. Arnold LE, Bozzolo DR, Hodgkins P, et al. Switching fromoral extended-release methylphenidate to the methylphen-idate transdermal system: continued attention-deficit/hyperactivity disorder symptom control and tolerabilityafter abrupt conversion. Curr Med Res Opin 2010; 26:129-37

26. Guy W. ECDEU assessment manual for psychopharma-cology. Rockville (MD): US Department of Health, Edu-cation, andWelfare, 1976: 218-22. DHEW publication no.:ADM 76-338

27. DuPaul GJ, Reid R, Power TJ, et al. ADHD rating scale –IV: checklists, norms, and clinical interpretation. NewYork: Guildford Publications, 1998

28. Bennett KJ, Torrance GW. Measuring health status prefer-ences and utilities: rating scale, time trade-off, andstandard gamble techniques. In: Spilker B, editor. Qualityof life and pharmacoeconomics in clinical trials. 2nd ed.Philadelphia (PA): Lippincott-Raven, 1996: 575-86

29. Torrance GW. Measurement of health state utilities foreconomic appraisal. J Health Econ 1986; 5: 1-30

256 Lloyd et al.


30. McGough JJ, Wigal SB, Abikoff J, et al. A randomized,double-blind, placebo-controlled, laboratory classroomassessment of methylphenidate transdermal system inchildren with ADHD. J Atten Disord 2006; 9: 476-85

31. Conners KC. Conners’ rating scales – revised: technical man-ual.NorthTonawanda,NY:Multi-Health Systems Inc., 2007

32. Lloyd AJ, van Hanswijck de Jonge P, Doyle S, et al. Healthstate utility scores for cancer-related anaemia through societaland patient valuations. Value Health 2008; 11 (7): 1178-85

33. Kind P, Dolan P, Gudex C, et al. Variations in populationhealth status: results from a United Kingdom nationalquestionnaire survey. BMJ 1998; 316: 736-41

34. Matza LS, Secnik K, Rentz AM, et al. Assessment of healthstate utilities for attention-deficit/hyperactivity disorder inchildren using parent proxy report. Qual Life Res 2005; 14:735-47

35. Kostopoulou O. The transient nature of utilities and healthpreferences. Med Decis Making 2006; 26: 304-6

Correspondence: Dr Andrew Lloyd, Director, Oxford Out-comes Ltd, Seacourt Tower, WestWay, Oxford, OX2 0JJ, UK.E-mail: [email protected]



Documents

Estimation of Utilities in Attention-Deficit Hyperactivity Disorder for Economic Evaluations