Transcript Development of a Lung Cancer Natural History Model

CT Screening for Lung Cancer
vs. Smoking Cessation:
A Cost-Effectiveness Analysis
Pamela M. McMahon, PhD; Chung Yin Kong, PhD;
Bruce E. Johnson; Milton C. Weinstein, PhD;
Jane C. Weeks, MD, MS; G. Scott Gazelle, MD, MPH, PhD
Department of Radiology & Institute for Technology Assessment
Massachusetts General Hospital
Harvard Medical School
Key Question

Modeling studies suggest that CT screening may
decrease lung cancer mortality

Several randomized trials will report mortality
endpoints in the next few years

If trials show evidence of benefit from CT
screening, would it be a good value relative
to other cancer control interventions?
Methods

Existing lung cancer simulation model to simulate
6 cohorts of individuals in multiple scenarios

For each scenario, predict total costs and total
quality-adjusted life years (QALYs) for the cohort



Calculate incremental cost-effectiveness ratios


costs in 2006 US $
costs & QALYs discounted at 3% annually
ICERs; defined as Δcost/ΔQALY
ICERs compared to benchmark of $100k/QALY
Lung Cancer Policy Model (LCPM)

Two versions – we used single cohort LCPM


population LCPM replicates US trends, 1975-2000
Model synthesizes available data


smoking histories from de-identified US survey data
observational studies, cancer registries (SEER)

Validated (single arm screening study)

Affiliate of the NCI
consortium
http://cisnet.cancer.gov/
Features of the LCPM

Microsimulation of individual life-histories


aggregated to cohort (population) statistics
Underlying natural history model

5 lung cancer cell types and benign nodules

Smokers face increased risks of death from
competing causes (e.g., CVD, COPD, others)

Screening biases and mortality reduction from
screening are predicted

based on model inputs (program characteristics)
Lung Cancer Policy Model
Schematic
General
population
Follow-up
Dead
Diagnosis
& Staging
Treatment
& Survival
Lung Cancer Policy Model
Schematic
Screening
General
population
Follow-up
Dead
Diagnosis
& Staging
Treatment
& Survival
Natural History Model

Risks of lung cancer depend on


accumulated smoking exposure
age, sex and birth cohort

Cancers grow (Gompertz) and can metastasize

Clinical staging, treatment modeled explicitly


can be varied to evaluate management strategies
Unobservable natural history parameters are
estimated through extensive model calibration
SEER
+ LCPM
Year
Incidence per 100,000
Interventions Compared

No intervention

Screening with helical CT

Smoking cessation alone

Combined CT screening/smoking cessation

Interventions modeled as one-time & annual
Inputs Relevant to the
Analysis

Cessation rates



Program characteristics




background annual cessation = 3%
effectiveness (1-year abstinence) = 4% to 30%
eligibility (age, pack-yrs, time since quitting),
adherence
number and frequency of screens, CT performance,
cost
follow-up protocol, +/- radiation risk
Costs


SEER-Medicare, CPT codes, wholesale prices
patient and caregiver time costs
Projected costs and effects –
base case (perfect adherence)
Results:
white males age 50 in 1990
$300 cessation with 16% abstinence at 1 year
Results
ICERs (in $1,000s/QALY)
Summary:
1) Combined interventions provided most benefit to most individuals but
yielded ICERs over $100,000/QALY (vs. cessation alone)
2) CT alone was dominated, regardless of cessation effectiveness
*Scenario shown on previous plot; all in cohorts of males aged 50
Additional Sensitivity Analyses

Including radiation risk for new lung cancers


 ICER for annual CT screening vs. no screening by
14% (70 year old men) to 85% (50 year old women)
Additional influential program characteristics


lower screening adherence increased ICERs
‘stricter’ eligibility for screening reduced ICERs for
annual screening but none were below $100K/QALY
Conclusions

Screening with helical CT costs more but
provides fewer benefits than cessation alone

Combined screening + cessation provides
benefits to more individuals but costs more than
$100,000/QALY (vs. cessation alone)

Results are dependent on model assumptions


model simulates guideline care
analyses are currently limited to whites

data on smoking histories and lung cancer incidence by single year,
which are needed for calibration, were not available for minorities
Acknowledgements


National Cancer Institute

R01 CA97337 + Supp (Gazelle)

R25 CA92203 (Gazelle)

R00 CA126147 (McMahon)
American Cancer Society

117494-RSGHP-09-148-01-CPHPS (Gazelle)

Colleen Bouzan, Angela Tramontano

CISNET lung cancer investigators