Upload
others
View
6
Download
0
Embed Size (px)
Citation preview
The health economics of genomic
sequencing
Helseøkonomikonferansen 2019 i Bergen, 27th May 2019
Dr James Buchanan
[email protected] | @jbuchanan_ox | https://healtheconomicsandgenomics.com
Health Economics Research Centre, Nuffield Department of Population Health, University of Oxford, Oxford, UK
Plan
1. Scientific background
2. Current genomic testing context in the UK
3. Existing health economics evidence base
4. Economic evaluation challenges
5. Two applied studies
a) Microcosting study of genome sequencing in cancer and rare diseases
b) DCE investigating stakeholder preferences for genome sequencing in
inherited cardiovascular disease
Background
Genetics and genomics
What is a gene?
• Genes are defined sections of DNA that are stored within cells in your body and which carry instructions for particular traits
• Everybody has around 20,000 genes. Your complete set of genes is your genome
What is genetics?
• The study of genes, their functioning and composition, and the way in which traits and conditions are passed between generations
What is genomics?
• The study of genomes, and the inter-relationships between all genes, in order to identify their combined influence on health and disease
What are genetic tests?
• Targeted to specific genes of interest
• e.g. newborn screening for cystic fibrosis
What are genomic tests?
• Provide information across whole genome; can identify multiple genetic changes
Genome sequencing
2003 – first human genome sequenced
2008 – next generation sequencing
approaches enter the research setting
• Genome sequencing / Exome sequencing /
Targeted panels
• Sequencing is quick, sensitive and
conducted at depth
• Multiple genetic changes can be detected
simultaneously vs genetic tests (often
targeted at specific genes)
• This genomic information can inform
diagnosis / prognosis / clinical management
e.g. in cancer and rare diseases
Genome sequencing in the UK
World leader
2008 onwards – genomic tests increasingly used in research setting
2013 – 100,000 Genomes Project launched
• Aim: sequence 100,000 whole genomes from patients with a rare disease or
cancer
• Completed in December 2018
October 2018 – NHS Genomic Medicine Service launched
• UK Government aims to sequence 5 million genomes in 5 years
What about cost-effectiveness?
Is genome sequencing cost-effective? If so, for which conditions?
Health economic evidence base in the UK to support this decision is limited
There is a Genomics England Clinical Interpretation Partnership (GeCIP) for Health Economics
• Unfunded / little work done to date
Some initial unpublished calculations performed for target conditions to estimate potential cost savings
No evidence on value
The international evidence base
Evidence on costs
No evidence that the cost of exome sequencing is falling over time
Limited evidence that the cost of genome sequencing is decreasing
Limited information on how cost estimates were generated
Many estimates either reflected commercial prices or were hypothetical
threshold costs
The lowest estimates were also the worst quality estimates
Evidence on health outcomes
Most common outcome measure was diagnostic yield
• Not widely accepted by health technology assessment agencies
Few studies reported health outcome measures
recommended for economic evaluations
• e.g. survival, quality of life
• Only 2 studies presented data on QALYs
Evidence on cost-effectiveness
First author
(Year)
Country Investigation Comparison Disease(s) Type of
economic
evaluation
Outcome measure Economic evaluation
results
Bennette
(2015)
USA Genome and
exome
sequencing
Returning
incidental findings
or not
Cardiomyopathy,
colorectal cancer,
healthy individuals
CUA QALYs £32,187-£82,623 per
QALY gained
Buchanan-
Hughes (2015)
UK Bacterial
genome
sequencing
Current testing
pathway
Urinary tract infections CUA QALYs Genome sequencing
dominated by current
practice
Sagoo (2017) UK Exome
sequencing
Current testing
pathway
Variety of conditions CEA Number of positive
diagnoses
£2,230-£3,213 per
additional diagnosis
Soden (2014) USA Genome and
exome
sequencing
Current testing
pathway
Paediatric
neurodevelopmental
disorders
CEA Number of diagnoses N/A (cost threshold
analysis)
Schofield
(2017)
Australia Exome
sequencing and
gene panel
Current testing
pathway
Childhood-onset
muscle disorders
(suspected congenital
muscular dystrophy or
nemaline myopathy)
CEA Number of diagnoses Cost-saving
Stark (2017) Australia Exome
sequencing
Three strategies
of integrating
exome
sequencing into
the current
testing pathway
Paediatric suspected
monogenic disorders
CEA Number of diagnoses £1,030-£3,830 per
additional diagnosis
Tsiplova
(2016)
Canada Genome and
exome
sequencing
Chromosomal
microarray
Paediatric autism
spectrum disorder
CEA Number of diagnoses £13,912-£106,590 per
additional diagnosis
Van
Nimwegen
(2017)
Netherlands Genome and
exome
sequencing
Current testing
pathway
Paediatric neurological
disorders
CEA Number of diagnoses Cost saving to £8,319
per additional diagnosis
Preferences (1)
Preferences (2)
11 studies identified
• 6 DCEs, 1 Contingent Valuation, 1 TTO, 3 Mixed methods
Marshall (2016):
• Ranking exercise to determine what people are willing to pay for genome sequencing information
• 38% would not pay for actionable genomic information, and 3% would pay more than $1,000
• 55% would not pay for genomic information for which medical treatment is currently unclear, and 7% would pay more than $400
Regier (2015):
• Preferences for the return of secondary findings from NGS
• This information is valued, but WTP depends on type of finding
Relevance? Secondary findings not a major focus of large scale sequencing projects
Health economic challenges
Analytical approach
Measuring costs
Measuring outcomes
Measuring effectiveness
Health economic challenges
Genomic test timing is critical – standard testing practice evolves
continuously
There are no national pricing tariffs for genomic tests
Disease-specific and preference-based outcome measures are limited
Capturing information on personal utility is important, but difficult
Effectiveness data for genomic interventions are challenging to incorporate
into standard health economic analyses
Oxford Microcosting Study
Oxford microcosting study
Aim: estimate the cost of using genome sequencing to identify
pathogenic variants in cancer or rare disease cases, using the
Illumina HiSeq 4000
Microcosting approach
• Costs assigned to individual resource use to generate aggregate costs
Study conducted in Oxford Molecular Diagnostics Centre
(NHS laboratory)
Costing based on annual throughput of 399 samples
Data collected from June 2016 to December 2017
Methods and data (1)
Collected information on resource use (equipment, staff, consumables) using questionnaires/interviews, then attached unit costs to generate an overall cost
Costed all steps in genome sequencing pathway, from sample reception to reporting and archiving
Same steps for cancer and rare diseases
Methods and data (2)
Unit cost data provided by laboratory staff or equipment suppliers
Equipment costs discounted at 3.5%
All salaries inflated by 20% to incorporate National Insurance and
Superannuation
Total costs inflated by 20% to account for overheads
All data assumed to be stored for 5 years
All parameters varied in one-way sensitivity analysis
All costs are 2016 values
Costs were calculated at the case level and also per genome:
• Cancer case size: 2 samples (tumour and germline)
• Rare disease case size: 3 samples (proband and both parents)
Microcosting results
Cost category (% of total cost before overheads)Total
Equipment Consumables Staff Overheads
Cancer (A)£694(12%)
£4,126(72%)
£880(15%)
£1,140£6,841
£3,420/genome
Rare diseases (B)£1,042
(18%)
£4,022(68%)
£811(14%)
£1,175£7,050
£2,350/genome
Difference (B-A) £348 -£105 -£69 £35 £209
0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
Pro
port
ion o
f to
tal te
st c
ost
s
Cancer
Rare diseases
Two-way sensitivity analysis (1)
Joint changes in annual throughput and consumable costs for genome sequencing in rare diseases
(results expressed as the cost per case)
£0
£2 000
£4 000
£6 000
£8 000
£10 000
£12 000
0 200 400 600 800 1 000 1 200 1 400 1 600 1 800 2 000
Co
st p
er c
ase
Throughput
Consumable cost = £100
Consumable cost = £500
Consumable cost = £1,000
Consumable cost = £4,023 (base case)
Annual throughput of 399 (base case)
Two-way sensitivity analysis (2)
Joint changes in annual throughput and consumable costs for genome sequencing in rare diseases
(results expressed as the cost per genome)
£0
£500
£1 000
£1 500
£2 000
£2 500
£3 000
£3 500
£4 000
0 200 400 600 800 1 000 1 200 1 400 1 600 1 800 2 000
Co
st p
er g
eno
me
Throughput
Consumable cost = £33
Consumable cost = £167
Consumable cost = £333
Consumable cost = £1,341 (base case)
Annual throughput of 399 (base case)
Microcosting conclusions
Our estimates are generally lower than existing estimates in the literature
Limitations – costs are sequencer/country specific
The costs of sequencing are yet to meet the desired £1,000/$1,000 per
genome figure when testing is performed on relatively small numbers of
patients with cancer or a rare disease in a single centre with modest
throughput
High throughput – commensurate with a national-scale facility – combined
with bulk discounts on consumable costs will likely have the greatest
impact on the overall cost of sequencing going forward
Is this likely?
DCE in inherited cardiovascular disease
Background
Patient access to genomic tests often strictly controlled
Preferences of healthcare professionals may influence the translation of genomic tests into clinical practice
Case study: genomic testing in inherited cardiovascular disease (CVD)
Several genomic testing approaches are options in inherited CVD
• Genome sequencing / Exome sequencing / Panel tests
No consensus on the best option
We conducted a DCE in health professionals in UK who order or who have ordered genetic or genomic tests for patients selected for mutation analysis in the context of inherited CVD
Aims:
• To understand the relative preferences of health professionals for the different attributes of genomic testing
• To consider how these preferences might impact on test uptake in the UK NHS
Attributes and levels
Test attribute
Possible levels for each testing alternative
Genome sequencing Exome sequencing Cardiac panel testGenetic testing not
indicated
Ability of the test to
identify pathogenic
mutations
Pathogenic mutation is identified in 20 out of every 100 cases
Pathogenic mutation is
identified in 0 out of every
100 cases
Pathogenic mutation is identified in 30 out of every 100 cases
Pathogenic mutation is identified in 40 out of every 100 cases
Pathogenic mutation is identified in 50 out of every 100 cases
Ability of the test to
identify variants of
unknown significance
Variant of unknown significance is identified in 10 out of every 100 casesVariant of unknown
significance is identified in 0
out of every 100 cases
Variant of unknown significance is identified in 20 out of every 100 cases
Variant of unknown significance is identified in 30 out of every 100 cases
Test cost
£1,000 £500 £150
£0£3,000 £1,500 £300
£5,000 £2,500 £450
£7,000 £3,500 £600
Quantity of counselling
received
40 minutes 10 minutes
0 minutes50 minutes 20 minutes
60 minutes 30 minutes
Disclosure of secondary
findings
No secondary findings disclosed
No secondary findings disclosedSubset of well-characterised secondary findings disclosed
Choice question
Experimental design
Modelled the probability that health professionals would select a specific
genomic test
Used model averaging approach in Ngene
Designs generated for both parts of the choice tasks – combined to create
overall d-efficient design
• Design for second part of choice task given weighting of 0.05
• Best estimate: only 5% of respondents would select the opt-out
Sampling and survey administration
Target respondents: healthcare professionals who order / have ordered
genetic and/or genomic tests for patients with a suspected inherited
cardiovascular disease
Sampled from two populations
• GeCIP in Cardiovascular Disease (N=200)
• UK Association for Inherited Cardiac Conditions (N=350)
Piloted in both populations
Survey conducted online
Respondents asked to rank attributes before/after completing choice tasks
Information collected on respondent characteristics
Data analysis
Evaluated observed uptake and existence of dominant preferences
Used mixed logit regression analysis
• Analysis 1: Evaluated responses to first part of choice question
• Analysis 2: Used responses to part two of choice question if respondent selected opt-out (only undertaken if >5% selected opt-out)
Evaluated impact of attributes/levels on test uptake
Also:
• Evaluated predictive power
• Calculated willingness-to-pay (WTP)
• Estimated marginal rates of substitution (MRS)
• Predicted future uptake
• Explored heterogeneity
Respondent characteristics
37 respondents
Key characteristics:
• Male (65%)
• Aged 35-44 years (41%)
• Cardiologists (40%) or clinical geneticists (31%)
• Saw one patient with suspected or confirmed inherited cardiovascular disease
every working day
• Most (95%) had previously ordered a genetic or genomic test
• Most (95%) thought that genomic tests for CVD patients should be made
available in the NHS
• Median time to complete DCE: 15 minutes
Responses to choice questions
57% selected exome sequencing in a practice question in which this was
the ‘best’ option
Panel test chosen most often (66% of all choices)
• Genome sequencing 18%, exome sequencing 14%, opt-out 1%
Five respondents (14%) always selected the choice alternative that
identified the most pathogenic mutations
Four respondents (11%) selected the panel test in every choice question
• These 4 respondents also selected panel testing in the practice question
Mixed logit
Attribute Β-coefficient SE Lower CI Upper CI PWillingness-to-
payLower CI Upper CI
Random parameters
ASC_WGSMean 3.194 2.700 -2.097 8.486 0.237 £486.27 -£198.45 £1,170.98
SD 0.386 0.758 -1.099 1.871 0.610 - - -
ASC_PANELMean 4.851 2.071 0.791 8.911 0.019 £738.52 £300.39 £1,176.65
SD 4.254 1.196 1.910 6.599 0.000 - - -
PATHOGENICMean 0.768 0.238 0.301 1.235 0.001 £116.90 £99.14 £134.65
SD 0.797 0.290 0.229 1.365 0.006 - - -
UNKNOWNMean -0.213 0.083 -0.375 -0.050 0.010 -£32.38 -£48.22 -£16.53
SD 0.304 0.110 0.087 0.520 0.006 - - -
COSTMean -0.007 0.002 -0.011 -0.002 0.003 - - -
SD 0.015 0.005 0.005 0.025 0.003 - - -
Fixed parameters
ASC_WES Mean 5.879 2.944 0.109 11.649 0.046 £894.94 £280.22 £1,509.66
COUNSEL Mean -0.097 0.043 -0.181 -0.012 0.025 -£14.72 -£24.30 -£5.15
SECONDARY Mean -0.052 0.474 -0.981 0.876 0.912 -£7.95 -£148.17 £132.26
Pseudo R2 0.24
Respondents willing to tolerate a 36% increase in VUS if the mutation detection
rate increases by 10%
Model predicts 99% of choices made by respondents
Test uptake
Test
Test attributes and levels Utility scoreProbability of
uptake
Pathogenic mutation identified
in X out of every 100 cases
VUS identified in X out of every
100 casesCost Counselling
Secondary findings
disclosedMean SD Mean SD
Genome sequencing 50 55 £5,000 60 mins Subset -18.5 30.8 0.1% 0.6%
Exome sequencing 45 35 £2,500 60 mins Subset 9.0 15.2 1.4% 4.1%
Panel testing 40 15 £300 30 mins None 33.9 15.9 98.1% 6.2%
None 0 0 £0 0 mins None -0.1 0.0 0.4% 2.3%
Conclusions
Healthcare professionals in this clinical context have strong preferences for
genomic testing, if the alternative is no testing at all
Uptake of genomic testing is more likely if the pathogenic mutation rate increases,
fewer variants of unknown significance are identified, or if tests decrease in cost
Respondents prefer panel testing to genome or exome sequencing
Either scepticism or lack of awareness of the benefits of genome and exome
sequencing, or awareness of the limitations of these tests in this context
• E.g. more VUS are identified using genome or exome sequencing, but few of these are
likely to convert to pathogenic mutations
Caveats: specific clinical context / single hypothetical scenario
Tentative conclusion: uptake of genome and exome sequencing might be limited if
these tests have a high yield of VUS
Where next?
1. How much do genomic tests cost?
• Need more microcosting studies in different settings and at scale
• Need to better understand all the costs that patients incur across their clinical
pathway, before and after testing
2. What are benefits of genome sequencing in terms of health outcomes?
• Survival and quality of life data are required
• Must collect data that can feed into current health technology assessment
processes
3. Is genome sequencing cost-effective?
• Urgent need for economic evaluations of different applications of genome
sequencing
Acknowledgements
Current and former* members of
the genomics team at the Health
Economics Research Centre
• Lars Asphaug*
• Sarah Briggs
• Ana Cruz
• Brett Doble*
• Patrick Fahr
• Jilles Fermont*
• Liz Morrell
• Laurence Roope
• Katharina Schwarze*
• Apostolos Tsiachristas
• Sarah Wordsworth
Microcosting study co-authors
• Pavlos Antoniou
• Carme Camps
• Helene Dreau
• Jilles Fermont
• Steve Harris
• Samantha Knight
• Erika Kvikstad
• Alistair Pagnamenta
• Melissa Pentony
• Niko Popitsch
• Anna Schuh
• Katharina Schwarze
• Jenny Taylor
• John Taylor
• Mark Tilley
• Sarah Wordsworth
DCE study co-authors and
collaborators
• Edward Blair
• Elizabeth Ormondroyd
• Jenny Taylor
• Kate Thomson
• Hugh Watkins
• Sarah Wordsworth
Takk!
Health Economics Research Centre, Nuffield Department of Population Health, University of Oxford, Oxford, UK
Extra slides
One way sensitivity analysis
Most changes in staff-related variables had no effect on test costs
HiSeq 4000 sequencing machine was the most expensive piece of
equipment (£474,373, annual maintenance £55,641)
• Varying the cost of the sequencer +/- 50% changed test costs by +/- 3-4%
• Half of the sequencer cost per cancer or rare disease case (£279 for cancer
cases and £418 for rare disease cases) is the annual maintenance cost
Changing the duration of data archiving had no effect on test costs
Reducing the family size for rare disease cases to 2.4 reduced test costs
by 20% to £5,650
Qualitative work
Literature review → Long-list of 16 potential attributes
Interviews with healthcare professionals (clinicians, laboratory
scientists, genetic counsellors)
• Ranked attributes by importance
• Five attributes selected
• Also discussed testing alternatives and attribute levels
Attribute rankings
Attribute Before completing choice questions After completing choice questionsMean score (SD) Ranking Mean score (SD) Ranking
Ability of the test to identify pathogenic mutations
1.0 (0.0) 1 1.0 (0.2) 1
Test cost 2.7 (0.9) 2 2.5 (0.9) 2Quantity of counselling received
3.4 (1.0) 3 3.5 (0.9) 4
Ability of the test to identify variants of unknown significance
3.6 (1.1) 4 3.4 (1.1) 3
Disclosure of secondary findings
4.4 (0.8) 5 4.5 (0.7) 5