Transcript Slide 1
Rescuing Clinical Trial Data For Economic Evaluation
Paul Kind, Ph.D.
Jan van Busschbach, Ph.D.
Frank de Charro, Ph.D.
Overview of the workshop
Frank de Charro: Introduction Jan van Busschbach: How to shoot straight at the goal Paul Kind: How to score with more subtle combinations
The problem
• Many clinical studies do not generate outcomes data that are suitable for economic evaluation • They may include condition-specific measures but these were generally not designed for the type of analysis required • Such studies may include measures of health-related quality of life as secondary endpoints
Economic evaluation
• Cost Effectiveness Analysis Measure of effectiveness can be anything as long as scaling properties are ok • Cost Utility Analysis Measure of effectivenss: QALYs
EQ-5D: Descriptive System
• Classification of health states using
5 dimensions
mobility self care usual activity pain / discomfort anxiety / depression • 3 problem levels for each (none / some / extreme) • defines a total of 3 5 = 243 health states Self-care Mobility Pain Health state Mood Usual activities
EQ-5D questionnaire
Tick one box for each group of statements.
Mobility
I have no problems in walking about I have some problems in walking about I am confined to bed
Self-Care
I have no problems with self-care I have some problems washing or dressing myself I am unable to wash or dress myself
Usual Activities
I have no problems with performing my usual activities (e.g. work, study, housework, family or leisure activities) I have some problems with performing my usual activities I am unable to perform my usual activities
Pain/Discomfort
I have no pain or discomfort I have moderate pain or discomfort I have extreme pain or discomfort
Anxiety/Depression
I am not anxious or depressed I am moderately anxious or depressed I am extremely anxious or depressed
1 0 3 2 State A : 1 1 2 2 3 3 2 1 0
Population TTO weights
State B : 1 1 3 2 2
Scoring EQ-5D health states
0.4
0.3
0.2
0.1
0 0.25
State A 0.36
0.11
State B Difference
The challenge
To devise methodologies that can be used to convert outcomes data collected in clinical studies into a form that have the necessary attributes to support economic evaluation (CUA)
Rescuing economic analysis
Clinical analysis/report Secondary data Clinical study Primary data Economic analysis / report
Cost-effectiveness Cost-utility
A word of caution
• Quality adjustment is one of the most important outcome characteristics • Indirect approaches involve uncertainty and diminish the potential to differentiate between a treatment and its alternativess • Cost effectiveness will be judged taking into account uncertainty and using probabilistic models • So crosswalks are second best but better than no coverage of utility at all
The valuation of disease specific health states
• Jan J. v. Busschbach, Ph.D.
– Erasmus MC – Institute for Medical Psychology and Psychotherapy – www.busschbach.nl
• Sildenafil (Viagra)
The effects of Sildenafil in terms QALY’s is complicated
• QALYs are measured with standardised and validated quality of life questionnaires • EQ-5D or HUI III were not included – Not sensitive for erectile dysfunction?
Clinical outcomes
• Gradations erectile dysfunction – were chosen as clinical outcomes • Measured with the IIEF – International Index of Erectile Function • Primary end points: Question 3 and 4 – Ability to attain an erection • Example: During intercourse I am sometimes able to penetrate – Ability to maintain an erection • I can almost never maintain the erection during intercourse after penetration
Clinical, disease specific outcomes
Baseline End of treatment 4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Placebo Sildenafil Placebo Sildenafil
Question 3 IIEF Question 4 IIEF Goldstein et al., N Engl J Med, 1998
How to convert clinical outcomes into QALYs ?
• 2 questions 5 answer levels = 25 health states – Why not value the 25 these health states with Time Trade-Off ?
• 169 subjects of the general public • Valued the 25 health states with TTO – Individual administration within groups sessions – Validation of procedure in students (group versus individual)
25 Erectile Dysfunction States
• During intercourse I am sometimes able to penetrate • I can almost never maintain the erection during intercourse after penetration
TTO values have logic structure
1
utilities (TTO)
0,95 0,9 0,85 0,8 0,75 0,7 0,65 0,6 0,55 0,5 almost never/ never a few times sometim es most times
(IIEF 4) maintenance of erections after penetration
almost always/ always almost always/always most times
(IIEF 3)
sometimes
frequency of
a few times
penetration
almost never/never
4.5
4.0
3.5
3.0
2.5
2.0
1.5
1.0
0.5
0.0
Transferred clinical outcomes into QALYs
Baseline End of treatment
On the basis of Goldstein et al., N Engl J Med, 1998
Placebo Sildenafil Placebo Sildenafil
Question 3 IIEF Question 4 IIEF
Baseline End of treatment 1.00
0.95
0.90
0.85
0.80
0.75
0.70
Placebo Sildenafil
Queston 3 and 4 transformed to QALY-weights
QALY league table
Intervention
GM-CSF in elderly with leukemia EPO in dialysis patients Long transplantation End stage renal disease management Hart transplantation Didronel profylase PTA with Stent Breast cancer screening
Viagra
Treatment of congenital anorectal malformations
$ / QALY
235,958 139,623 100,957 53,513 46,775 32,047 17,889 5,147
5,097
2,778
Disease specific utilities are not equal to generic utilities
Health y No complains • Only the disutility of the specific disease is valued • Generic and specific utilities are not on the same scale – Generic top anchor: absence of any impairment – Specific top anchor: absence of specific impairment • Co morbidity might still be present Death All complains
How to interpret disease specific utilities
• Value of life years “traded off” in TTO differ – Healthy subject: – Sick subject: 1 life year is 1.0 QALY 1 life year is 0.5 QALY – Life years of healthy persons are more worth than those of sick • Overall health states influence disutility – 20% trade off at 1.00: disutility = 0.20
– 20% trade off at 0.80: disutility = 0.16
– 20% trade off at 0.60: disutility = 0.12
• Raw disease specific trade-off overestimated gains
Specific utilities should be corrected for average morbidity
• Solution: multiplicative model • Multiply disease specific value with average value • Values have to be multiplied by average value for age group.
– For instance in IPSS • male age 55-64: overall QoL utility: 0.81
• Most severe BPH: 0.87
• Male age 55-64 with most severe BPH: 0.81 x 0.87 = .7047
• Maximum gain reduces from – Raw score 1.00 - 0.87 = 0.13
– Adjust score 0.81 - 0.70 = 0.11
– 15 % reduction
Rue of thumb
• Overestimated CE-ration by 15% using specific utilities – Proposed by Fryback & Lawrence, MDM 1997 • For not completely the same problem… • …for own health states, not imaginable health states
Conclusion (1)
• We validated the IIEF and the IPSS for the use in economic appraisal – Now, IPSS and IIEF has QALY-weights • Many other applications possible – (health states of…) diabetic foot ulcers • Advantage – High sensitive disease specific measures for QALY-analysis – No need for generic instrument • Disadvantages – Not directly compatible with generic utilities….
– ± 15 % correction needed
Overestimation?
• Does the focus on the disease makes the disutility to high?
Crosswalks: recalibrating
Paul Kind
Visiting Professor
University of Uppsala, Sweden
Principal Investigator
Outcomes Research Group Centre for Health Economics University of York England
Recalibration – the task
• Source assumed to be a clinical / condition specific (sensitive) measure • Format – Summary score / Index – Subscale scores / dimension scores – Items (all or selected) • Target assumed to be a generic index weighted using social preferences • Task – to recalibrate source in terms of target
Recalibration strategies - direct
• Derive direct estimates of social preferences for source index • May require simplification of complex descriptive system • Will have implications for time and resourcing • May conflict with instrument developer agenda
Recalibration strategies - indirect
• Multiple solutions linking all or part of source instrument with the target index (directly or indirectly) Strategy A Estimate target index from A1 source index A2 sources subscales A3 source items Strategy B Estimate target dimension/levels from B1 source subscales B2 source items
Strategy A1
• 25 item condition sensitive instrument with widespread usage in its therapeutic field • Yes/no answers coded to 1/0 • All items assumed equal weight • Summary index • General population survey of circa 1,000 yielded parallel observations with EQ-5D
18.000
19.000
20.000
21.000
22.000
23.000
24.000
25.000
A mean estimated 0.000
observed 0.943
0.970
1.000
0.931
0.949
2.000
3.000
4.000
5.000
6.000
7.000
8.000
9.000
10.000
11.000
12.000
13.000
14.000
15.000
16.000
17.000
0.893
0.873
0.900
0.817
0.872
0.820
0.820
0.852
0.743
0.795
0.780
0.747
0.793
0.712
0.564
0.626
0.929
0.908
0.888
0.868
0.847
0.827
0.806
0.786
0.766
0.745
0.725
0.704
0.684
0.664
0.643
0.623
0.761
0.275
0.602
0.582
0.562
0.541
0.521
0.500
0.480
0.460
EQ 5D’ = 0.9696 – 0.0204* A
1 0.8
0.6
0.4
0.2
0 -0.2
0 2 4 6 8 10 12 14 16 18 20 22 24
Issues
• Number of observations across “severity” range • Subgroup impact – Age / gender • Significant factors but small effect • Regression on mean observations – Why not micro level ?
• Less good fit • Tricky / messy business • May not significantly improve estimation – EQ5D’ = 0.9419 – 0.0163
Strategy A3
EORTC QLQC-30
• EORTC QLQC-30 is a generic measure of health-related quality of life (HrQoL) in cancer. Version #3 consists of 28 items with a 4-category response and 2 further items (general health and quality of life) are coded on a 7-point response category scale (see selected items below).
• Responses are converted into corresponding numeric scores that may be summed to produce a total score. However, QLQC-30 cannot be used in cost-effectiveness analysis because it is not standardized on a value scale where full health = 1 and dead = 0.
EORTC QLQC-30
selected items
Data
• Baseline observations from a previously reported study of 177 patients with pancreatic cancer were available for analysis. HrQoL in these patients had been assessed by self-report using both the QLQ-C30 and EQ-5D measures.
• Additional baseline data on patients included their Karnofsky Performance Scale rating
Methods
• The first 28 QLQC-30 items were dichotomised (not at all = 0 ; quite a bit to very much = 1) and these items, together with the uncoded response to item 29 (general health) were entered in a stepwise linear regression in which EQ 5D index was the dependent variable.
Results
• The results of this regression analysis are given in Table 1, showing that only 6 of the QLQC-30 items proved to be significant.
• The r 2 of 0.490 equates with levels seen in other such calibration studies. Correlation between observed EQ 5D index and estimated values was generally high. However, correlation between observed EQ-5D index and estimated value amongst female patients is higher than for male patients in the sample (r = 0.725 vs 0.662). Mean differences across all patients was -0.001
• Figure 1 shows the scatterplot of observed EQ-5D and the value derived for each patient using the 6-item QLQC-30 model
Table 1 : Coefficients from stepwise OLS regression model
Std. Error Significance Unstandardized Coefficients 0.633
(Constant)
.071
.000
Q29 Overall health QX3 trouble with short walk QX26 physical family life impact QX5 help with dressing washing QX20 difficulty concentrating QX11 trouble sleeping
0.047
-0.124
-0.082
-0.167
-0.102
-0.086
.013
.031
.031
.047
.033
.032
.001
.000
.009
.000
.002
.007
Figure 1 : Observed and estimated values for EQ-5D index 1.0
.8
.6
.4
.2
0.0
-.2
0.0
EQ-5Dindex .2
.4
.6
.8
1.0
Figure 2 : Mean observed and estimated EQ-5D index values for categories of Karnofsky Performance Scale 1.0
.8
.6
.4
.2
0.0
N = 29 * 29 62 * 62 KPS at baseline 44 * 44 28 * 28 5 * 5 EQ-5Dindex index based on QLQC-30 items
Have a nice walk
But also Try to keep going straight if possible
www.euroqol.org