How do we evaluate the cost-effectiveness of new medicines? What value do we place on effective drugs? Prof. Ken Paterson explores the challenging area of health economics and how we judge whether we can afford new treatments in a cash-limited health service.
Glomerular Filtration and determinants of glomerular filtration .pptx
The Dismal Scientist: the price of everything, the value of nothing
1. The Dismal Scientist!
The Price of Everything, the Value of Nothing
Ken Paterson
RMCSG/GSMS Joint Meeting
10 January 2013
2.
3.
4.
5. Health Technology Assessment
► Why do we need it at all?
► What is it?
► How is it actually done?
Is it about science or just about money?
Can it really assess costs, benefits and value?
Is it too complex for ‘real people’ to grasp?
► What can prescribers (and patients) add?
► Are ‘they’ all heartless bean-counters?
…or is it the way forward?
6. Increasing Pressures in all
Health-Care Systems
► Population demography
Aging population in almost all countries
Increasing obesity and physical inactivity
► Patient/Public expectation
More possibilities to intervene beneficially
“a pill for every ill!”
Lifestyle drugs/disease mongering
► Changing environment for new medicines
Mature market place
7. Pharmaceutical Market 2013
► Three decades of major advances
► Effective medicines for most common diseases
Often now ‘generic’ agents at low cost
Most needs met (at least to some extent)
► New medicines of two main types
Extend choice in existing crowded areas
New options in niche areas of unmet need
► Do the benefits justify the (opportunity) costs?
8. Medicines Licensing
► Safety
► Quality
► Efficacy (v placebo)
► Not comparative efficacy (or effectiveness)
► Not ‘place in therapy’
► Not ‘value for money’
► Licensing important but limited
9. Health Technology Assessment
► Provides a logical framework for decisions
► Has to compare many options –
Different disease areas
Different interventions
Different patient groups
► Has to be as objective and dispassionate as
possible
…but still with a human face
► Inevitably produces ‘winners’ and ‘losers’
10. HTA - How is it Actually Done?
► Liraglutide – a new GLP-1 agonist
Usually more expensive than exenatide
Possibly better efficacy/tolerability
► A new option after 1 or 2 existing therapies
Usually more expensive than TZD or DPP-4
Different efficacy and tolerability profile
Different ‘added value’ – eg weight
► Is it a good use of limited NHS money?
11. Clinical Issues for HTA
► S, Q and E covered by licensing
► Comparative efficacy tested
► Comparative safety tested
…includes tolerability
► Likely effectiveness in ‘real-world’ clinical
practice tested
…will it do what it says on the tin?
► Overall, is it ‘value for money’?
12. Liraglutide Study Programme
► LEAD 1 – on SU, compare L to TZD
► LEAD 2 – on MF, compare L to SU
► LEAD 4 – on MF/TZD, compare L to glargine
► LEAD 5 – on MF/SU, compare L to glargine
► LEAD 6 – on MF/SU, compare L to exenatide
► Which is/are the most relevant comparisons?
No comparison to DDP-4 inhibitor (‘gliptin’)
13. How Do We Measure Health Gain?
► 2 main domains
Quantity of life (= survival)
Quality of life
► QALY (quality-adjusted life year) gets both
14. How To Measure Quality of Life
► Utility from 1 (perfect health) to 0 (death)
► Visual analogue scale
► Generic questionnaire
EQ-5D has 200 health states with utility for each
► Preference–based measures
Time trade off
Standard gamble
► An inexact science – ?what would be better
15. Utility Value Examples
► Diabetes, no complications 0.814
► MI, 2 years after event 0.736
► Congestive heart failure 0.633
► Painful neuropathy 0.624
► Stroke, 2 years after event 0.545
► Haemodialysis 0.490
► BMI (per kg/m2 above 25) 0.006-
16. Trial v Lifetime Benefit
► Clinical trial shows short-term effects on
surrogates
HbA1c
Blood pressure
Lipids
Weight
► HTA needs to see the consequent
outcomes
Mortality
Morbidity
► Health economic ‘modelling’ the key
17. Modelling in Diabetes
► LOTS of great epidemiological data
… including outcome of interventions
► LOTS of complexity
Many surrogates
Many outcomes
►Macrovascular events
►Microvascular complications
► How can we put all this together?
19. The Actual Modelling
User sets simulation
conditions
Generate baseline
population
No Any patients
to run?
Stop Yes
Time horizon
Yes Reached?
No
Screening
ACEI
treatment
LASER
treatment
Statin
treatment
Aspirin Neuro- Foot ulcer, Retino- Macular Nephro-- Hypo- Keto- Lactic Non-spec.
treatment MI Angina CHF Stroke PVD Cataract
pathy amputation pathy edema pathy glycemia acidosis acidosis mortality
Specific Specific Specific Specific Specific Specific Specific Specific
mortality mortality mortality mortality mortality mortality mortality mortality
Overall annual
survival
Time counter
advances
Update simulation
data
20. CORE Diabetes Model
►CORE Diabetes Model ►UKPDS Outcomes Model
Angina Ischaemic heart disease
Myocardial infarction
Myocardial infarction
Heart failure
Heart failure
Stroke
Stroke Blindness
Peripheral vascular disease Renal failure
Diabetic retinopathy and blindness Amputation
Macular oedema ► First occurrence only
Cataract
Hypoglycaemia
Ketoacidosis, lactic acidosis
Nephropathy and end-stage renal disease
Neuropathy
Foot ulcer and amputation
21. CORE Model – How Reliable?
CORE Diabetes Model values
66 validation analyses
100
R2 = 0.9222
y = 1.0187 x
80
60
40
20
0
0 20 40 60 80 100
Published study values
22. Basic Model Structure
Cohort Economics
Treatment Clinical (costs and
(baseline)
utilities)
Data processing
(calculating annual risks
for 15 different complications)
The results
(predicting the clinical outcomes
and associated HE results:
incidence rates,
LE, QALE, costs, and more)
23. Simple Markov Model
Healthy Ill Dead
User sets simulation
conditions
Generate baseline
population
No Any patients
to run?
Stop Yes
Time horizon
Yes Reached?
No
Screening
ACEI
treatment
LASER
treatment
Statin
treatment
Aspirin Neuro- Foot ulcer, Retino- Macular Nephro-- Hypo- Keto - Lactic Non-spec.
treatment MI Angina CHF Stroke PVD Cataract
pathy amputation pathy edema pathy glycemia acidosis acidosis mortality
Specific Specific Specific Specific Specific Specific Specific Specific
mortality mortality mortality mortality mortality mortality mortality mortality
Overall annual
survival
Time counter
advances
Update simulation
data
24. Costs to NHS by Health
Healthy Ill Dead
Cost: € 25 Cost: € 2500 Cost: € 0
Not only medicines costs but ALL costs
from NHS budgets
A costly medicine may offset its cost by
producing savings elsewhere
25. Quality of Life
Healthy Ill Dead
Cost: € 25 Cost: € 2500 Cost: € 0
QoL: 1.0 QoL: 0.5 QoL: 0.0
26. Over Time – nothing changes…
Healthy Ill Dead
Cost: € 25 Cost: € 2500 Cost: € 0
QoL: 1.0 QoL: 0.5 QoL: 0.0
27. …or you can move on!
Healthy Ill Dead
Cost: € 25 Cost: € 2500 Cost: € 0
QoL: 1.0 QoL: 0.5 QoL: 0.0
28. …and we know the chances!!
30%
Healthy Ill Dead
Cost: € 25 Cost: € 2500 Cost: € 0
QoL: 1.0 QoL: 0.5 QoL: 0.0
15% 20%
80% 50%
5%
Chances are obviously dependent on time -
these are the chances per cycle length (1 year)
29. How Does the Model Run?
► We know all the clinical issues
…and chances of progression/improvement
► We know the impact of changes in -
Weight
HbA1c
BP
Lipids
► … on chances of progression/improvement
► We know how the interventions (liraglutide
and comparators) affect these surrogates
30. Monte-Carlo Simulation!
► We know our diabetes population
► Let’s invent 1000 patients
… representative of ALL Scottish patients
► Let’s treat them with liraglutide or comparator
► Let’s run the model until they all die (~20 years)
► Let’s see the costs and health gains with
liraglutide and with the comparator
For each patient calculate the extra cost and extra
QALYs (or less cost or QALYs)
► Let’s see what the overall cost-per-QALY looks
like
31. The Actual Markov Model
User sets simulation
conditions
Generate baseline
population
No Any patients
to run?
Stop Yes
Time horizon
Yes Reached?
No
Screening
ACEI
treatment
LASER
treatment
Statin
treatment
Aspirin Neuro- Foot ulcer, Retino- Macular Nephro-- Hypo- Keto - Lactic Non-spec.
treatment MI Angina CHF Stroke PVD Cataract
pathy amputation pathy edema pathy glycemia acidosis acidosis mortality
Specific Specific Specific Specific Specific Specific Specific Specific
mortality mortality mortality mortality mortality mortality mortality mortality
Overall annual
survival
Time counter
advances
Update simulation
data
34. Liraglutide 1.2mg v Exenatide
£10,000
£5,000
£0
-1 -0.5 0 0.5 1
∆ Costs (£)
-£5,000
-£10,000
∆ QALY (years)
Most patients are cheaper with most getting
health-gain – liraglutide DOMINANT
35. Liraglutide 1.8mg v Exenatide
£10,000
£5,000
£0
-1 -0.5 0 0.5 1
∆ Costs (£)
-£5,000
-£10,000
∆ QALY (years)
Most patients cost more but get more health-gain
– median cost-per-QALY £15,581
36. Sensitivity Analysis
► A series of ‘what-if’ tests
► Allows exploration of doubt in the model
Remove or add individual factors (eg mortality)
Allow for ‘real-world’ rather than clinical trial
Use different utility (QoL) values
► Shows what are the key assumptions
We can then test these with real clinicians
We can focus our final judgment on these
► Cost-per-QALY is not set in stone!!
37. Liraglutide 1.2mg Cost-Effectiveness
► LEAD 1 – on SU, L v TZD £10,751
► LEAD 2 – on MF, L v SU £23,598
► LEAD 4 – on MF/TZD, L v glarg £7,801
► LEAD 5 – on MF/SU, L v glarg £8,847
► LEAD 6 – on MF/SU, L v exen dominant
38. Liraglutide 1.8mg Cost-Effectiveness
► LEAD 1 – on SU, L v TZD £17,394
► LEAD 2 – on MF, L v SU £43,369
► LEAD 4 – on MF/TZD, L v glarg £14,923
► LEAD 5 – on MF/SU, L v glarg £17,777
► LEAD 6 – on MF/SU, L v exen £15,581
39. Cost-Effectiveness Threshold
► <£20K – YES; >£30K – NO
► Plucked from thin air – no ‘scientific’ basis
Industry (?patients) think it should be higher (!)
Economists think it should be lower (?£15K)
► Provides a level of fairness even if wrong!
… but links poorly to other initiatives
►Motorway barriers
►Automated train signalling
►Airport security
► Worthy of greater discussion/debate
40. What Can Prescribers Add?
► Information on real-world practice
… what is/are the best comparator(s)?
… how are these patients looked after?
… what is the extent/nature of unmet need?
► Comment on the health economic case
Not the detail but the clinical assumptions
Is the model capturing all benefits (and ADRs)?
Economists (pharma and HTA!) good at maths
and models, not clinical practice
41. The Larkin Tests!
► The Poke of Chips test
Drug X outcomes…..
Drug X + docetaxel outcomes…..
► The Larkin Test
“If this medicine does very little and costs £5000
per year, why is the cost per QALY £10?”
► Sense-checking always crucial!!
42. What Can Patients Add?
► Real insights into their experience of diabetes
► Detail on the problems of current therapy
Tolerability
Convenience
► Potential patient sub-groups for new therapy
… beyond pure clinical trial data
… based on vignettes/anecdote at times
► Main role is in finely-balanced decisions
43. How Good Are New Medicines?
► Manufacturers’ assessments of lifetime
health gain for 256 medicines
27% NO health gain
20% >0 but ≤0.1 QALY
25% >0.1 but ≤0.5 QALY
15% >0.5 but ≤1.0 QALY
13% >1 QALY
Median 0.14 QALY; mean 0.59 QALY
44. HTA - Implications for
Prescribers
► Good prescribing cannot ignore cost
► Cost-effectiveness is not about saving money
► Why would you prescribe outwith guidance?
You don’t care!
►Why not? - it’s not your money!!
There are patient specific benefits not captured
►Are you sure? Why would they be omitted if real?
There are reasons to ‘push the boat out’
►Good reasons? How would you justify if asked?
45. Do We Need to Redefine
‘Value’?
► Licensing based on ‘risk:benefit’
► Benefit is the impact on the disease
…easily measured in clinical trials
► Value is the impact of ‘risk:benefit’ on the
patient
…not easily measured in clinical trials
May well need ‘modelling’, use of surrogates etc
May not be measured in £££/€€€/$$$
► Health economic modelling or value
modelling?
46. Anti-Disease to Pro-Patient
► Anti-cancer
► Anti-retroviral
► Anti-infective
► Anti-fungal
► Anti-diabetic
► Anti-hypertensive
► Anti-inflammatory
► Where is the patient?
47. Medicine for Cancer or Patient?
•Two medicines for the same cancer -
outcome at 8 months
– Medicine A - 30% reduction in primary
tumour size; mean 12.5 kg weight loss,
mean loss of 1.5 in performance status;
mean 28 days as in-patient
– Medicine B - 30% increase in primary
tumour size; no change in mean weight; no
change in performance status; mean 4 days
as in-patient
48. The Paradigm Shift
► NOT ‘how does the medicine impact the
disease’
RECIST, viral load, CRP etc
► BUT “how does the medicine impact the
patient with the disease’
Perception of health
Functional abilities
Quality of life
► The real value of a medicine (or anything else)
49. Value Assessment in Healthcare
► Probably a necessary ‘evil’
The worst way apart from all the others!
► Principles of health economics are sound
Clinically based
Far from ‘facile’, ‘simple’ or all about price
Aims to capture all benefits and true value
► Money wasted in NHS helps no-one
… and deprives other patients of real benefits
► Health economists (and friends) are human -
they and their subject are not DISMAL!
50. Thomas Carlyle
“That there should one
man die ignorant who
had capacity for
knowledge, this I call a
tragedy!”
“The greatest of
faults, I should say,
is to be conscious of
none!”
51. Oscar Wilde
“It is always a silly
thing to give advice,
but to give good advice
is fatal!”
“Always forgive
your enemies -
nothing annoys
them so much!!”