Epidemiology and Oncology Translational Research in Clinical Oncology October 24, 2011 Neil Caporaso, MD Genetic Epidemiology Branch, Division of Cancer Epidemiology and Genetics, National Cancer Institute

1
Epidemiology and OncologyTranslational Research
in Clinical Oncology October 24, 2011Neil
Caporaso, MDGenetic Epidemiology
Branch,Division of Cancer Epidemiology and
Genetics, National Cancer Institute
2
NIH National Cancer Institute Division of
Cancer Epidemiology and Genetics Genetic
Epidemiology Branch
National Institute of Health Includes many
institutes.
We are INTRAMURAL 85 are extramural
Cancer ETIOLOGY
Other Branches focus on Nutrition, Hormones,
Infection, Occupation, Statistics, Radiation
3
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
4
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
5
What are the goals of epidemiology ?1. Identify
the causes of cancer2. Quantify risks3.
Identify risk groups4. Understand
mechanisms5. Public health and health
services6. Identify syndromes
6
Epidemiologist as a crusher of dreamsQuestion
asked of the consulting epidemiologist What
is the p value?What the epidemiologist is
thinking..1. Your study design is?2. Your
controls came from.?3. Did you consider bias,
confounding?4. What was the original hypothesis?
(data dredging)5. Did you consider power?6. Did
you validate your marker?etc.Epidemiologist is
helpful when a question involves the population
(as opposed to an individual, organ, cell, etc.)

7
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
8
Domain of epidemiology Epidemiology causes of
health and disease in human populations epi
(upon) demos (the people) logia (talk about)
An OBSERVATIONAL science (like astronomy,
evolutionary biology)-Contrast with
experimental-Investigator does NOT get to pick
who is exposed or unexposed -Free-living people
make choices about participatingpossible
BIAS-Study of individuals with and without
disease (contrast with Clinical Research)
9
RisksA measure of the strength of the
relationship between the risk factor and the
cancerRelative (RR)Involves comparison of 2
groups of individualsSo, if tobacco has a RR10
for lung cancer,smokers are 10-fold more likely
to get lung cancer than non-smokers.RR is a
ratio Absolute (AR) Involves risk over a
specific time periodOften more clinically
relevant.So, there might be a 20 lifetime risk
of lung cancer in a smokerAR is a difference
Odds ratio (OR) is similar to RR but noteOR is
calculated in a case-control study, RR from a
cohort studyProper interpretation of an OR
requires the outcome event to be RARE

10
Prevalence Total number of cases in population
at a given timeDISTINCT from incidence (NEW
cases ascertained during a time
period)Prevalence better for LONG duration
diseases, incidence for SHORTLifetime prevalence
and Point prevalenceKEY POINT- low prevalence is
associated with FALSE POSITIVE TESTS of people
diagnosed with cancerIncludes those cured and
those living with the diseasegt 10 million
Americansare cancer survivors
11
Epidemiologists emphasize prevention
Rationale-Effective (think polio, smallpox,
smoking cessation, clean water, HPV)-Cheap
(compared to late stage interventions)-Public
health orientation-Eliminate disease at the
sourceDownsides-Often requires time to
demonstrate effectiveness -Less dramatic than
treatment-Harder to see all the disease you have
prevented-Less positive political impact-Fewer
Nobel PrizesPrimary directed to susceptibility
stageExample Needle exchange to prevent AIDS,
HPV vaccineSecondary directed to subclinical
stageExample Screen for cervical cancer with
Pap SmearTertiary directed to clinical
stageExample Treat diabetic retinopathy to
prevent blindness
12
The most common question epidemiologists get!My
grandmother smoked all her life. She drank
heavily, her exercise was the TV remote,she
never used a seat belt, she had bacon and
buttered toast for breakfastShe out lived all
her doctors.. She drank shots on her 90th
birthdayThe race is not to the swift or the
battle to the strong, nor does food come to the
wise or wealth to the brilliant or favor to the
learned but time and chance happen to them all.
(Ecclesiastes)Deterministic vs. Probabilistic
13
Epidemiologists are ethically prohibited
fromdoing experiments on people So, we observe
large populations and seehow their outcomes
relate to what people do (i.e., smoke, drink,
eat, etc.)This argument was used for years by
tobacco companiesto deny evidence linking
cigarettes and cancerObservational vs.
Experimental

14
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
15
Cancer mortality map
16
GIS Geographic patterns of disease and exposure
via satellite. Examples, used to estimate
nitrate, pesticide levels
17
SEER
18
SEER Surveillance, Epidemiology, and End Results
(SEER) Program 26 of US populationincidence
and survival, patient demographics, primary tumor
site, tumor morphology and stage at diagnosis,
first course of treatment, and follow-up for
vital status comprehensive source of
population-based information

19
CANCER RATESThese are RATES not numbers of
eventsKEY DIFFERENCERates take into account age
and size of population
20
Cancer Incidence Rates, All Sites Combined, All
Races, 1975-2000
21
Men cancer rates
75 increase due to PSA screening
22
Cancer Incidence Rates for Women, US, 1975-2000
Rate Per 100,000
Breast
Colon rectum
Lung
Uterine corpus
Ovary
Age-adjusted to the 1970 US standard
population. Source Surveillance, Epidemiology,
and End Results Program, 1973-1998, Division of
Cancer Control and Population
Sciences, National Cancer Institute, 2001.
23
Cancer incidence rates
24
Cancer death rates
Why are cancer death rates leveling off?
25
Lung cancer death rates
..because the most common cause of cancer death
is declining
26
Men cancer death rates
27
Women cancer death rates
28
Children and cancer
29
Childhood Cancers (lt 14 ys) 
Incidence 8,600 new cases/yr 12,400
(0 19 ys) Mortality 1,500 deaths/yr
2,300 (0 19 ys) rates ? 50 since 1973
Etiology -- poorly understood

30
Cancer in men
31
Cancer in women
32
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
33
What are the general risk factors for
cancer?Increasing ageEnvironmental
factorsGenetic factorsCombinations of the
above!
34
Most Cancer is due to the EnvironmentDramatic
differences in cancer rates by geography and over
time are only compatible with extrinsic
environmental causesEstablished by a vast body
of descriptive, ecological, and analytical
epidemiology
35
International Variation in Cancer RatesType of
cancer H/L highest lowestMelanoma 155 Austra
lia JapanNasopharynx 100 Hong
Kong UKProstate 70 US (Blacks) ChinaLiver
50 China CanadaCervix 28
Brazil IsraelStomach 22 Japan KuwaitLung
19 US (Blacks) IndiaColon 19 US
(Whites) IndiaBladder 16 Switzerland IndiaPan
creas 11 US (Blacks) IndiaOvary 8
Maori (NZ) KuwaitBreast 7 Hawaii
IsraelLeukemia 5 Canada India
36
Causes of Cancer Deaths Environmental pollution,
Infectious agents, Lifestyle, Alcohol use,
Occupational factors, Medicine, Radiation,
Genetic susceptibility, other unknown causes
Tobacco 30-35
Diet 30-35
Other 30-35
37
Copper smelter, Montana
Copper Smelter, Montana
38
Lung cancer mortality rate in Xuan Wei is among
the highest in China
County-specific female lung cancer mortality
rates (per 100,000, 1973-75)
39
Indoor Air Pollution in China
40
Skull With Cigarette van Gogh
41
How do you prove a cause?1. It should confer
high risk2. It should be consistent3. Dose
response4. Cause occurs first!5. Biology makes
sense
42
Lung cancer increases with cigarette smoking
43
Smoking leads to lung cancer
44
Squamous tumors
45
Basal cell carcinoma
46
Lung cancer and smoking cessation
47
Per-Capita Consumption of Different Forms of
Tobacco in The U.S. 1880-2003
Data Source USDA
48
Environmental Tobacco Smoke (ETS)In 1981 2
studies suggested never-smoking women spouses of
smokerswere at higher risk then spouses of
non-smokers (Hirayama, Trichopoulos)NRC
ReportNonsmoking spouses have 30 increased risk
(1.3-1.7 OR)¼ of cases in non-smokers
attributable to smoking or 2-3Estimated 3000
deaths per yearSurgeon General Report (1986) and
EPA Review (1992) concurred ETS
classified as Class A human carcinogenMethods
critiqueSmall sample sizeExposure
misclassificationOmission of confoundersMetanaly
ses conclude that ETS (both Workplace and at
home) is a significant risk factor, e.g. Law,
1997SummaryEvidence implicating ETS is
consistent with model-based estimates and
suggests dose-response extends to lowest
exposures, i.e. no threshold
49
Tobacco and public health major cause of
preventable morbidity mortality 1/5
US deaths (450,000 USA, 3M world/y) 10 M
tobacco deaths/yr (2030, WHO) 30 of cancer, 8
sites, all difficult to treat tobacco related
disease costs Medicare/ Medicaid gt 10B/yr
each In spite of widespread knowledge of the
health consequences of smoking - rates in US
adolescents are stable or increasing - declines
in adults- leveled off - individual smoking
cessation difficult
50
What are alcohol-associated cancers?OralPharyn
xEsophagusLarynxLiver
51
Radiation

• Ionizing
• Non Ionizing
• Ultraviolet
• Electromagnetic

52
Ionizing RadiationLeukemia (AML, but not
CLL)BreastLungThyroidHead and neck cancer
53
Partial list studies implicating cancer and
Ionizing RadiationType of XRT Study Cancer
ImplicatedA-Bomb Japan Breast, Leuk, Gastric,
ThyA-Bomb Marshall Island ThyroidMedical Breas
t/Mastitis BreastMedical Hemangioma Breast,
ThyroidMedical Hodgkins Breast, lung,
ThyroidMedical TB-Flouroscopy Breast Radionucl
ides Thorotrast Leukemia, Liver
(Th-232)Radionuclides Spondylytis Bones
(Ra-224)Occupation Radium Dial
painters BoneOccupation Rad Technicians Leukemia
Occupation Chernobyl Cleanup ?Environmental In
door radon Lung
54
Non-Ionizing Radiation (UV/sun)1. Basal
cell2. Squamous cell3. Melanoma

55
Excessive sun tanning
56
What are some dietary risk factors?High
fat Colon, breastHigh calories UterineLow
fiber ColonMicronutrients Lung (?)Diet
contaminents Liver
57
Diet and lung cancerMany questions..1.
Failure of nutrient based interventions(ATBC
and beta carotene)2. Role of processedvs.
traditional food3. Obesity paradox (and
smoking)4. Food?/nutrients?How to best
aggregate consumed items to identify risk or
protection?5. Meat and vegetable consumption
58
Higher frequency of fresh red and processed meat
intake increased lung cancer risks
Fresh red meat Processed
meat
p-trend lt0.001
Lam et al, 2009, Cancer Res.
59
Viruses and cancer
60
Newer viral hypothesesVirus Human Cancer
(hypothesized)HCV hepatocellular
cancer NHLEBV NPC Hodgkins
lymphoma leiomyosarcomaKSHV
(HHV8) Kaposis sarcomaHPV-16, -18, -33,
-39 Vulvo-vaginal cancer Anal
cancer Penile cancer oropharyngeal
cancerPolyomavirus Merkel cell virus/
CLL?HIV non-hodgkins lymphoma
61
Bacteria and Stomach Cancer

• Helicobacter pylori increases risk of stomach
  cancer

62
HP-associated Disease (US)
63
We find no association of H. Pylori with lung
cancer
64
Cloned Familial Tumor Suppressor Genes
Retinoblastoma RB1 13q14 1986Wilms
tumor WT1 11p13 1990Li-Fraumeni
syndrome p53 17p13 1990Neurofibromatosis
1 NF1 17q11 1990Neurofibromatosis
2 NF2 22q12 1993von Hippel-Lindau
VHL 3p25 1993Familial melanoma
1 p16 9p21 1994Familial breast
1 BRCA1 17q21 1994 Familial breast
2 BRCA2 13q12 1995Basal cell nevus
PTC 9q22 1996
65
Occupational Exposure FactorsArsenic (Blot and
Fraumeni 1994, Lubin 1981)Sheep dip and vineyard
workersAsbestos (Doll 1955, Blot 1978, Blot and
Fraumeni 1981)Textile workers, miners, millers,
insulation, shipyard, cementChyrsotile (most
common commercial- least carcinogenic)15-20 year
latent period- late stage effectSynergy with
tobaccoChloromethyl ether (Pasternack
1977)Small cell carcinoma in chemical
workersChromium (Machle 1948)Mustard gas (Wada
1968)Nickel (Doll 1977)PAHs (IARC 1983)Radon
(Samut 1989)Uranium miners, household radon,
medical radiationSilica (IARC 1987)
66
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
67
gaps on the ENVIRONMENT sideFor many cancers,
risk factors are unknown?For cancers where
general cause, is understood, individual
susceptibility is poorly understoodHow G and E
work in concert is poorly understood.Some
potential causes are poorly studied
Epidemiologists have responded byMolecular and
integrative epidemiologyAdvanced
technologyConsortia
68

• Chronic Lymphocytic Leukemia
  Most common leukemia
  of Western world.
• 30 of adult leukemia in USA
• Less frequent in Asia and Latin America.
• Male to female ratio is 21.
• Median age at diagnosis is 65-70 years.
• No extrinsic environmental causes known
• Family history is the most important risk factor

69
All Cancer is due to the Genetic changesAll
cancer cells exhibit changes in theirDNA that
are passed on and maintainthe malignant
phenotype
70

• GETTING ORIENTED
• distinctions for genetic studies
• 1. Germline or Somatic
• (inherited or in the tumor)
• 2. Family or Population
• (rare or common)
• 3. Candidate or Agnostic
• (candidate gene study or GWAS)

71
Genetic Epidemiology

• Etiology, distribution, and control of disease in
  families and with inherited causes of disease in
  populations
• Includes
• family studies

72
To look for rare genes you need families
73
CDKN2A Mutations in Familial Melanoma

• CDKN2A -- major melanoma susceptibility gene
• Frequency of mutations varies in families
• 2 cases lt5
• 3 5 cases 20 24
• gt6 cases 50

74
Cloned Familial Tumor Suppressor Genes
Retinoblastoma RB1 13q14 1986Wilms
tumor WT1 11p13 1990Li-Fraumeni
syndrome p53 17p13 1990Neurofibromatosis
1 NF1 17q11 1990Neurofibromatosis
2 NF2 22q12 1993von Hippel-Lindau
VHL 3p25 1993Familial melanoma
1 p16 9p21 1994Familial breast
1 BRCA1 17q21 1994Familial breast
2 BRCA2 13q12 1995Basal cell nevus
PTC 9q22 1996
75
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
76
The lung cancer challenge.1- Drives overall
cancer mortality in the US and worldwide2-
Treatment and screening pose challenges3- Lung
cancer is paradigm for genetics of complex
disease4- Clearest example of environment and
gene in cancer5- The clearest example of a
genetically influenced behavior associated with
the leading public health problem in the world
77
Lung Cancer Risk and Family History
Family member OR (95
CI)Mother 2.11 (1.11-4.41)Father 1.37
(1.01-1.87)Sibling 1.53 (1.10-2.12Any family
member 1.57 (1.25-1.98)






78
Starter paradigm for identifying candidate
genes in lung cancer Smoking causes most lung
cancerCarcinogens in tobacco must be
metabolically activatedMetabolic alteration is
under genetic control

79
Processing is often under hereditary
control examples tobacco nicotine
(CYP2A6) aryl amines (NAT2) PAH (CYP1A1, GSTM1,
others) nitrosamines (CYP2A6/13, CYP2E1)

80
What were some challenges in finding genes
involved in common cancers using candidate
approach (pre-2007)?- type 1 error (false
positives)- population stratification -
multiple comparisons- inadequate power (type 2
error)design issues - failure to consider
gene-environmentfailure to consider
pathwaysfailure to consider genetic architecture
81
Categories of Cancer Causation The relationship
of genes to the environment
Environment -
Genes -
82
Best example of gene-environment in
cancer..NAT2 detoxifies aromatic amines
(acetylator phenotype)Aromatic amines are
bladder cancer carcinogensAromatic amines cause
high bladder cancer rates in dye workersAromatic
amines are major carcinogens in tobacco
smokeGWAS (rs1495741) confirms 2 decades of
candidate gene studies
83
Why GWAS (Genome Wide Association Study)?INITIAL
SKEPTICISM Collins/Gelertner wager in
2008STRONG DESIGN features High density SNP
genotyping arrays large patient samples with
detailed phenotype information minimize bias
from population substructure or variation in DNA
processing strict statistical thresholds for
significanceSTRONG FINDINGS agnostic.genes/path
ways implicated often previously unsuspected
links between diseases and pathways findings
outside transcription units ORs typically lt 1.5
GAPS not much epistasis or formal
gene-environment to date much heritability
unexplained mechanisms and precise functional
variants often unclear blind spots coverage of
many candidate genes, MAF lt 5, subgroups
84
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
85
Traditional epidemiology
Exposure
Disease Tobacco Lung
Cancer

86
Molecular epidemiology
exposure internal dose early biological
effect altered structure or
function early disease disease

87
Integrative epidemiology Behavior
exposure internal dose early biological
effect altered structure or
function early disease
diseaseOutcome

88
Integrative epidemiology Behavior
exposure internal dose early biological
effect altered structure or
function early disease disease
Outcome Instruments Fagerstrom Nicotine
DependencyDSM-IV Nicotine DependencyHospital
Anxiety and Depression Eysenck Personality
InventoryCESD- DepressionAttention Deficit
InventoryAttitudes and Knowledge about
SmokingIntention to Quit Smoking

89
Integrative epidemiologyQuestions from GWAS
perspective 4 categories of questionsHow are
genes related to?1. Disease (disease subsets
like histology) 2. Exposure (behavior, dose)3.
Markers of early/late disease (tissue/disease
markers)4. Outcome (survival, therapy efficacy
and side effects, px)

90
EAGLE website
91
Study DesignPopulation-basedCatchment's area 5
cities and 216 municipalitiesCases from 13
hospitals Controls randomly sampled from the
areaMatched by age, sex, and residence
92
Study description2000 incident lung cancer cases
and 2000 population-based controlsTest smoking
initiation in 700 never smokers and 600 former
smokers.Test smoking persistence with 700
current smokers and 600 former smokers
Participation rate, Cases85 Controls73
93
Why Population Controls ? 1. Gold standard2.
Representative of the population from which cases
derive3.Can calculate absolute rates4.Reduces
selection biasIMPLIESDefined population in
time and spaceSpecified eligibility and
exclusion criteriaDefined and high response
rate
94
To search for the 2nd (common) category of genes
you need large populations Sample Size
for Detecting Gene Effects
3000
2500
2000


1500
1000
500
0
0.0
0.1
0.2
0.3
0.4
0.5
95
Pilot studies participation rate 30 get a
Phone Survey 49 get a Invitation letter,
Follow-up by phone, In hospital, Advertisements,
Cash award, Home/hospital and Physicians
letter. 73 get a New interviewers, Physicians
call, Gas coupon, TV ads, New invitation letter,
Mayors letter and Toll-free phone line. Total
number of subjects in pilot investigations
156 Cases - 212 Controls Clinical data
99 Questionnaires 87 Biospecimens 97

96
Lung Cancer Case Control


97
EAGLE example molecular epidemiology
approachEpidemiology Biospecimensdoneness
module
98
EAGLEEnvironment And Genetics in Lung Cancer
Etiology PLCO Cancer Screening TrialProstate,
Lung, Colon, OvaryInitial design included 5800
subjects but we sought collaborators from
otherlung cancer studies to have additional
power to find genes..
99
3 regions identified in lung cancer GWAS
Broderick et al, Cancer Research , 2009
100
Major difference in chr 5 SNP by histology
101
What genes have we found?
1. TERT on
chromosome 5 a gene that maintains the telomeres
of chromosomes also found in other cancers and
lung diseases only contributes to ONE
histology! 2. CHNRA3 on chromosome 15 the
nicotinic receptor increases nicotine
dependency likely contributes to both lung
cancer and smoking!
102
A Population Perspective on Cancer1. Role of
epidemiology2. Epi-speak3. Epidemiologists
tools 4. What causes cancer?5. Why dont we
know about cancer causes?6. Why bother with lung
cancer if we know the cause?7. Why study design
matters? 8. What next?
103
QuestionsWhat causes lung cancer in
nonsmokers? 7th leading cause of death Why do
some successfully treated cancers recur?
operable lung cancer recurs in gt 50 What
causes cancers with no known exposure? CLL,
brain cancer, sarcomas ??? Where is the
hidden genetic variation? we cant find genes
for some cancers.. What is it about diet that
contributes to cancer? what aspect of diet is
really important?

104
What has molecular epidemiology contributed?3
examples1 HPV is the cause of 100 of
cervical cancer - prevention is possible
(vaccine)2 Cutting down on smoking is
ineffective - biomarker studies show levels of
carcinogens dont decline3. GWAS
studies (100 conditions) based on
biospecimen collections

105
Large integrative studies provide key
advantages- incorporate new technologies and
disciplines test diverse hypotheses lower
marginal costs (efficient) bring
interdisciplinary expertise to bear full
scientific value from large study platforms
contribute to consortiaLarge studies can
combine multiple domains to address key
questions..