Lynch syndrome – Economic evaluations - PowerPoint Presentation

Slide1

Lynch syndrome – Economic evaluations

Tristan Snowsill

Test Club 5

th

December 2017

Slide2

Lynch syndrome

Heritable

cancer syndrome

HeritableCaused by constitutional mutations affecting the DNA mismatch repair genes MLH1, MSH2, MSH6 and PMS2Autosomal dominant pattern of inheritanceIncomplete penetrance

Cancer syndrome

Carriers at increased risk of

Colorectal cancer

Endometrial cancer

Ovarian cancer

Plus others (gastric cancer, small bowel, …)

Cancer incidence at earlier age than gen. pop.

No premalignant signs of Lynch syndrome

Slide3

Lynch syndrome

Population prevalence (≈ incidence at birth)

Around 1 in 400

Estimated to cause

≈ 3% of colorectal cancer and endometrial cancerDisproportionately responsible for early-onset cancer (up to 8%)Cancer survival tends to be somewhat better than sporadic cancer

Slide4

Interventions

Surveillance

Risk-reducing surgery

Chemoprevention

Slide5

Diagnosis (affected by CRC/EC)

Limited use

Clinical and pathological features

Family history

Suggestive, but not diagnosticMicrosatellite instability (MSI) testingMMR immunohistochemistry (IHC)Excluding likely sporadic casesMLH1 methylation testingBRAF V600E mutation testing (CRC only)

Slide6

Diagnosis (affected by CRC/EC)

Diagnosis through identification of pathogenic constitutional mutation in or affecting MMR genes

Sequencing for short mutations

MLPA for genomic mutations (deletion, inversion, repetition of exons)Interpretation can be challenging

Slide7

Predictive testing

When a Lynch syndrome mutation is known in the family, can test blood relatives for mutation

Very accurate

Not very expensiveUsually use cascade testing

Slide8

Genetic counselling

Usually individuals are referred for genetic counselling prior to mutation testing

Helps them to understand

What is being tested for

What the risks areWhat options are availableCounselling also recommended following testingBut some suggest substantial streamlining

Slide9

(How) should we try to identify people with Lynch syndrome?

Two broad routes to diagnosis in a previously unidentified family

Referred after being affected by cancer (possibly reflex)

Referred due to strong family history of cancer

Population screening (with or without risk assessment) unlikely to be cost-effective compared to reflex or opportunistic testing

Slide10

Decision problems

Criteria

PenTAG

HTA report 2014

PenTAG NICE DAR 2016Current workPopulationEarly-onset CRCCRC (all ages)EC (subgroup by age)

Tests

IHC, MSI,

BRAF

,

MLH1

methylation

IHC, MSI,

BRAF

,

MLH1

methylation

IHC, MSI,

MLH1

methylation

Risk-reducing interventions

Colonoscopies

Risk-reducing surgery

Colonoscopies

Risk-reducing

surgery

Gynae

s

urveillance

Aspirin

Colonoscopies

Outcomes

Incremental cost-effectiveness ratio (ICER)

Costs

QALYs

Study design

Economic

model (individual patient sampling)

Economic model (Markov cohort)

Slide11

Model structure

Decision tree for affected individual

Leads to:

Lynch syndrome diagnosed

Lynch syndrome assumedAssumed sporadicThen individual patient sampling or Markov model used to predict long-term outcomesIf Lynch syndrome diagnosed then predictive testing offered to relativesIf Lynch syndrome assumed then relatives may be offered surveillance

Slide12

Summary of findings

PenTAG

HTA report 2014

Life expectancy for those with Lynch syndrome improved by over 1 yearIncreased use of colonoscopiesDecreased CRC incidence and mortality

All testing strategies cost-effective versus no testingIn fully incremental analysis MSI with BRAF was cost-effectiveIncreasing age threshold worsens cost-effectiveness

Slide13

Summary of findings

PenTAG

NICE DAR 2016

Testing still cost-effective versus no testing (except for direct mutation testing)Cost-effective strategy is IHC plus BRAF

and MLH1 methylation testing

Slide14

Current work

Cost-effectiveness of different testing strategies to identify Lynch syndrome in women with endometrial cancer

Pragmatic literature reviews for model parameters

Sophisticated(?) meta-analyses and parametric model fittingDecision tree and Markov model

Extensive exploration of uncertainty

Slide15

Tests with more than two outcomes

Microsatellite instability testing

Typically use panel of 5 markers

0 affected

 Microsatellite stable (MSS)1 affected  Low microsatellite instability (MSI-L)2-5 affected  High microsatellite instability (MSI-H)MMR immunohistochemistryPanel of four antibodies for MMR proteinsAbsent/weak/abnormal staining suggests deficiency in MMR systemBUT MSH6/PMS2 proteins break down if they cannot dimerize with MSH2/MLH1Intention to diagnose?

Slide16

Intention to diagnose

If individual has following pattern of IHC staining:

MLH1 absent

MSH2, MSH6 and PMS2 present

Is it a true positive result if the individual actually has an MSH2 mutation?What are the likely downstream tests?

Slide17

Bivariate meta-analysis

MSI testing

MMR immunohistochemistry

Slide18

Test failure

What are thresholds for considering tests to have failed?

MSI

Any markers non-evaluable?

No markers evaluable?IHCAny proteins non-evaluable?MSH6/PMS2 non-evaluable?

Slide19

Prevalence of Lynch syndrome

Assume age of cancer onset is Normally distributed given LS status:

For each study we have an age limit

and by Bayes’ rule we can relate the observed prevalence

to the counterfactual prevalence in an unselected cohort with the Bayes’ rule:

We fit a random effects model for the log-odds prevalence of LS:

This is all fitted using the Stata

ml

suite

 

Slide20

Gene distribution of LS in EC patients

Variation due to random chance or heterogeneity across studies?

Slide21

Incidence of colorectal cancer (fitting)

Slide22

Cumulative risk of colorectal cancer

Slide23

Plans

Current work

Manuscript drafted by Christmas/New Year

Send to people who may be interested in being co-authors or “internal reviewers” (former collaborators)Submit to journal (TBD) by end of February 2018

Next stepsCollaborate with Neil Ryan, Emma Crosbie, Katherine Payne in Manchester to integrate with their analyses based on PETALS study

Slide24

Thank you!

Any questions?