Slides for Course EPIB 671

1
McGill University Department of Epidemiology
Biostatistics Summer Session - 2008 Cancer
Epidemiology and Prevention Course EPIB
671 Course Coordinator Eduardo L. Franco,
Professor Departments of Epidemiology
Biostatistics and Oncology Director, Cancer
Epidemiology Unit (514-398-6032) E-mail
eduardo.franco_at_mcgill.ca
2
Course EPIB 671 CANCER EPIDEMIOLOGY AND
PREVENTION - 2008 Department of Epidemiology
Biostatistics, McGill University Eduardo Franco
(514-398-6032, eduardo.franco_at_mcgill.ca) http//ww
w.mcgill.ca/cancerepi/courses/caepisum/
Session Date Topics to be covered Articles
1 May 5 (Mon) Introduction, mechanisms of carcinogenesis, tumour biology, descriptive epidemiology
2 May 7 (Wed) Descriptive epidemiology (contd), causality, epidemiologic approaches and study designs Taubes
3 May 9 (Fri) Epidemiologic methods (contd) Gorey
4 May 12 (Mon) Causes tobacco, lifestyle Schiffman
5 May 14 (Wed) Causes Diet (200 pm Dr. Parviz Ghadirian) and occupational factors (340 pm Dr. Mark Goldberg)
6 May 16 (Fri) Causes infections
7 May 19 (Mon) Presentations by students on topics of their choice
8 May 21 (Wed) Prevention primary and secondary Giovanucci
9 May 23 (Fri) Prevention (contd) take-home exam, course evaluation Kyzas, Eaker
3
EPIB 671 (Summer Session) Files to download for
students registered in EPIB 671 Course
Description and Bibliography general information
about the course, its contents and bibliography
(not essential in class but be sure to read
before coming to class on Monday)EPIB671_CourseSc
hedule2007 .pdfCourseDescriptionSummer2007
.pdf Entire set of transparencies to be used
by Dr. Franco(needed in class to facilitate
note-taking)Slides-EPIB671-2007 .pdf
or,Slides-EPIB671-2007-condensed_BW .pdf
IARC-Monographs-Evaluation Summaries Appendix
material (Not essential but will be frequently
alluded to in class)IARC_Monographs .pdf
TaubesScience269-164s and TaubesLetters Article
by Taubes, Science 1995 and associated letters to
the editor (Please read it. It will be discussed
in class)TaubesScience269-164s .pdf
JNCI85-958 Article by Schiffman et al., JNCI
1993 (Please read it. It will be discussed in
class)JNCI85-958 .pdf Gorey-AJPH-1997
Article by Gorey et al., AJPH 1997 (Please read
it. It will be discussed in class)Gorey-AJPH-199
.pdf Chapter5-FrancoRohan Chapter 5 of
Franco Rohan textbook on the epidemiology of
cancer precursors. Contains supplemental
information concerning etiologic models and
measurement error (Not essential in class but if
you have a chance read it before the session on
methods, specially if you have not taken an
intermediate level epidemiology
course)Chapter5-Franco-Rohan.pdf .pdf
CaDetPrev-26-350 Review article by Franco et
al., 2002. Contains detailed information
concerning guidelines for cancer screening and
prevention (Not essential. Useful as reference
material to supplement the discussion during the
session on screening and prevention)CaDetPrev-26-
350 .pdf Chapter by Franco in the Encyclopedia
of Cancer 1997 Contains an overview of cancer
epidemiology and prevention. Its contents reflect
the general layout of the course (Not essential
but supplements the discussion on general
applications of epidemiology in cancer
research)EncyclopCancerChapter .pdf Review
article by Franco et al., Sem Cancer Biol 2004 on
causal relations in cancerSemCaBiol-14-413-2004.p
df .pdf Giovannucci-NEJM-1995 Article by
Giovannucci et al., NEJM 1995 (Please read it. It
will be discussed in class)Giovanucci-NEJM
.pdf Franco-JCE.pdf Article by Franco et al.,
JCE 1993. (Please read it. It will be discussed
in class).FrancoJCE .pdf OccupationalCancer.pp
t Occupational Cancer Lecture by Dr.
GoldbergGoldberg's Lecture 2007 .ppt Dr.
Ghadirian's Lecture 2007 Nutrition and Cancer
Lecture by Dr. Ghadirian.Ghadirian's
Lecture_2007 .pdf Summary of incidence
ratesGlobocan_2002_All_sites_Incidence .pdf
Article by Kyzas et al., JNCI 2005 (will be
discussed in class).JNCI_Kyzas_97_1043_2005
.pdf Article by Eaker et al. and Commentary by
Franco, PLoS-Medicine 2006PLoS-Med-Eaker-3-321-2
006 .pdfPLoS-Med-Franco-3-e48-2006 .pdf
4
Expanded Purview of Cancer Epidemiology

• Cancer surveillance burden of disease, incidence
  and mortality trends, cancer clusters
• Cancer risk assessing candidate etiologic
  factors
• Cancer prevention assessing the validity and the
  impact of chemoprevention and other preventive
  approaches
• Cancer screening assessing efficacy, comparing
  competing technologies
• Cancer survival assessing prognostic factors,
  determinants of quality of life in terminally ill
  patients

5
Milestones in Cancer Epidemiology

• Establishment of first tumour registries (1935,
  1943) and development of data quality standards
  by the IARC and IACR (1970s)
• Doll Hill (1950) Wynder Graham (1950)
  case-control approach to study cancer causes
  (cigarettes and lung cancer)
• Surgeon General's Report on tobacco and cancer
  (1964)
• WHO's IARC founded in 1965 major contributions
  CI5C and carcinogenicity monograph series
• Doll Petos report to the US OTA (1981)
• Emergence of molecular epidemiology (late 80's)
• Launching of mega-studies of screening (60's -
  80's) and diet (80's)
• Focus on precursor lesions as opposed to
  clinically invasive cancer (90s)
• Studies of SNPs and genome-wide association
  studies (2000s)

6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
Age-adjusted rates of death from lung cancer in
relation to smoking and asbestos exposure
Using a multiplicative scale Smoking
11.2 Asbestos 5.3 Expected RR with no
independent effects 11.2 x 5.3 59.4 Using an
additive scale Smoking (11.2 1) x 100
1010 Asbestos (5.3 1) x 100 430 Expected
with no independent effects 1440 or
equivalently, RR15.4
Adapted from Hammond et al., 1979 Beaglehole et
al. 1993
10
Caretaker Genes
Gatekeeper Genes
DNA repair Carcinogen metabolism
Cell cycle control Programmed cell death
Shields and Harris, 2000
11
(No Transcript)
12
(No Transcript)
13
From Cavenee, 1995
14
Cancer causation the Darwinian process Mel
Greaves Lancet Oncol 2002 3 24451 Clonal
evolution of a cancer. All cancers evolve by
Darwinian principles clonal proliferation,
genetic diversification within the clone, and
selective pressure enabling mutant subclones to
bridge the bottlenecks (such as anoxia,
restricted space and nutrients, apoptosis
imposition). Each colour in the figure represents
a cell (and its descendent clone) acquiring the
first (blue) or additional, sequential mutations.
Grey represents dying cells. This diagram greatly
simplifies the extensive genetic diversity,
complex population structure, and highly variable
dynamics of cancer clones. N, normal stem cells.
15
Adapted from Ruddon, 1995
16
Non-epidemiologic approaches used in assessing
the evidence concerning the carcinogenicity of a
suspected chemical, physical, or biological
exposure or its circumstances (Adapted from
Franco et al., SCB 2004)
Approach Type of scientific evidence Level of inference Type of study Features
Mechanistic Analogy Molecular structure Structure-activity relationships Useful to identify potentially carcinogenic compounds based their molecular similarity to known carcinogens
Toxicology Experimental DNA, cellular, organ In vitro short-term genotoxicity assays Rapid screening system for candidate compounds or exposures
Toxicology Experimental Organ, whole organism In vivo animal studies Provides proof of principle and insights into dose-response effects
Other supporting in vivo and in vitro data
relevant to evaluation of carcinogenicity can
also be used, particularly if they provide
insights into mechanisms of absorption,
metabolism, DNA binding or repair,
hormonally-mediated effects, genetic damage,
altered cell growth, loss of euploidy, cytopathic
changes, and related biological effects.
17
Epidemiologic approaches used in assessing the
evidence concerning the carcinogenicity of a
suspected chemical, physical, or biological
exposure or its circumstances (Adapted from
Franco et al., SCB 2004)
Type of epidemiologic evidence Level of inference Type of study Features
Observational Non-inferential, descriptive Case reports Suggestion of association
Observational Population Surveillance of incidence and mortality Documentation of baseline disease burden, exploratory hypotheses
Observational Population Ecologic (correlation or aggregate) studies Coarse verification of correlation between exposure and disease burden
Observational Individual Cross-sectional studies Correlation between exposure and disease (or marker) without regard to latency
Observational Individual Case-control studies Correlation between exposure and disease (or marker) with improved understanding of latency suitable for rare cancers
Observational Individual Cohort studies Correlation between exposure and disease (or marker) with improved understanding of latency suitable for rare exposures
Experimental Individual Randomized controlled trials of preventive intervention Most unbiased assessment of correlation between exposure and disease (or marker)
RCTs may target communities or providers as
units of randomly allocated intervention.
However, this is done for convenience of study
design in practical terms inference is at the
individual level.
18
Coverage of IARCs Cancer Incidence in Five
Continents Monographs
19
Estimated numbers of new cancer cases and deaths
in 2002 (Parkin et al., CA Cancer J Clin 2005)
20
Estimated numbers of new cancer cases and deaths
in 2002 (Parkin et al., CA Cancer J Clin 2005)
21
Age structure of developing and developed
countries
Male
Female
Developing
Developed
Proportion ()
Source IARC, 2000
22
(No Transcript)
23
Effect of Choice of Standard Population for
Age-adjustment
Gender Cancer Site Rate according to standard population Rate according to standard population Difference (US-World)
Gender Cancer Site US 2000 World 1960 Difference (US-World)
Males Prostate 177.6 117.7 50.9
Lung 82.1 51.5 59.4
Testis 5.6 5.1 9.8
Female Breast 137.1 99.0 38.6
Cervix 8.0 6.3 27.1
Vulva 2.4 1.5 56.7
Average age-adjusted incidence rates per 100,000
(1998-2002) in the US SEER program
24
ASIR (x 100,000), All sites except skin
non-melanoma top 10 and bottom 10 countries,
Males
(Source Globocan 2002)
25
ASIR (x 100,000), All sites except skin
non-melanoma top 10 and bottom 10 countries,
Females
(Source Globocan 2002)
26
ASIR (x 100,000), Liver carcinoma top 10 and
bottom 10 countries, Males
(Source Globocan 2002)
27
ASIR (x 100,000), Cervical cancer top 10 and
bottom 10 countries
(Source Globocan 2002)
28
Age-adjusted death rate (per 100,000 men)
Year
Age-adjusted death rates in the US (2000
population) Source American Cancer Society,
Surveillance Research
29
Age-adjusted death rate (per 100,000 women)
Year
Age-adjusted death rates in the US (2000
population) Source American Cancer Society,
Surveillance Research
30
Age-standardized (2000 US population) incidence
rates in areas covered by the US NCIs SEER
program. Source Ries et al., 2008 and previous
reports
31
Age-standardized (2000 US population) incidence
rates in areas covered by the US NCIs SEER
program. Source Ries et al., 2008 and previous
reports
32
Canada Incidence rates among men (age-adjusted
to the 1991 Canadian population)
Source Canadian Cancer Statistics 2008
previous ones
33
Canada Incidence rates among women (age-adjusted
to the 1991 Canadian population)
Source Canadian Cancer Statistics 2006
previous ones
34
(No Transcript)
35
(No Transcript)
36
(No Transcript)
37
(No Transcript)
38
From Armstrong and Mann, 1985
39
(No Transcript)
40
Design Layout of a Cohort Study
From Beaglehole et al., W.H.O., 1993
41
Design Layout of a Case-Control Study
From Beaglehole et al., W.H.O., 1993
42
(No Transcript)
43
Components of Etiologic Models in
Cancer Commonly Suspected Relations V1 and V2
candidate risk factor variables 1 and 2 O cancer
outcome Adapted from Franco et al., 2002
44
Components of Etiologic Models in Cancer Less
Suspected Mechanisms V1 and V2 candidate risk
factor variables 1 and 2 O cancer
outcome Adapted from Franco et al., 2002
45
(No Transcript)
46
Effect of measurement error in epidemiologic
studies
Parameter RR (exp-dis) Assumptions P(exp)20,
P(dis)2.5
Adapted from Franco and Rohan, 2002
47
Relative risks for associations between HPV and
cervical cancer in case-control studies of first
generation
NAH non-amplified DNA hybridization PCR
polymerase chain reaction
48
Cumulative incidence of SIL among women with a
normal Pap smear at entry (Local cytology)
HPV positive
HPV negative
Ludwig-McGill Cohort (August 1997)
49
Cumulative incidence of SIL among women with a
normal Pap smear at entry (Review cytology)
HPV positive
HPV negative
Ludwig-McGill Cohort (August 1997)
50
Features of Epidemiologic Study Designs
Features Ecologic Cross-sectional Case-control Cohort Randomized controlled trial
Study of rare outcomes Appropriate No Appropriate No (unless high risk population is targeted) No (unless high risk population is targeted)
Study of rare exposures Appropriate No No Appropriate Not applicable
Study of multiple outcomes Appropriate Appropriate No Appropriate Appropriate
Study of long latency No No Appropriate Inefficient Inefficient
Assessment of temporality Possible No Possible Yes Yes
Can measure incidence? No No Only if all cases identified Yes Yes
Weight of evidence Very low Low High Very high Highest
Types of biases Ecologic fallacy, confounding, detection, misclassification Selection, recall, confounding, misclassification Selection, detection, recall, confounding, misclassification Selection, detection, confounding, misclassification Misclassification, differential loss to follow-up
Study duration Very short Short Intermediate Long Long
Cost Very low Low High Very high Highest
Modified from Beaglehole et al. 1993
51
(No Transcript)
52
(No Transcript)
53
Criteria to Establish Causality (Hill, 1965)
Most important Experimental evidence Strength of
association Consistency Temporality Biologic
gradient Least important Coherence Plausibility
Analogy Specificity
54
(No Transcript)
55
OVERALL EVALUATION OF CARCINOGENICITY (Internation
al Agency for Research on Cancer, W.H.O.)

• Group 1 The exposure circumstance is
  carcinogenic to humans.
• Sufficient evidence of carcinogenicity in humans.
• Evidence less than sufficient in humans but
  sufficient in experimental animals and strong
  evidence that in exposed humans the agent acts
  through a relevant carcinogenic mechanism.
• Group 2A The exposure circumstance is probably
  carcinogenic to humans.
• Limited evidence in humans but sufficient in
  experimental animals.
• Inadequate evidence in humans but sufficient in
  experimental animals and strong evidence that in
  exposed humans the agent acts through a relevant
  carcinogenic mechanism.
• Group 2B The exposure circumstance is possibly
  carcinogenic to humans.
• Limited evidence in humans and less than
  sufficient evidence in experimental animals.
• Inadequate evidence in humans but limited
  evidence in experimental animals with supporting
  evidence from other relevant data.
• Group 3 The exposure circumstance is not
  classifiable as to its carcinogenicity to humans.
• Evidence inadequate in humans and inadequate or
  limited in experimental animals.
• Evidence inadequate in humans and sufficient in
  experimental animals but carcinogenic mechanism
  in animals does not operate in humans.
• Group 4 The exposure circumstance is probably
  not carcinogenic to humans.
• Evidence suggesting lack of carcinogenicity in
  humans and in experimental animals.

56
(No Transcript)
57
Corroboration of Epidemiologic Findings

• A golden rule?
• Provides the necessary confidence for public
  health action
• Provides the knowledge base that serves as
  foundation for mechanistic studies

58
Corroboration of Epidemiologic Findings

• The downside epidemics of repetition
• Newly discovered associations tend to lead to
  successive attempts at replicating the original
  findings
• Strong or moderate associations become clear with
  few replications
• Weak associations can only be examined with a
  large and diverse base of studies
• False associations may lead to a frivolous
  barrage of studies infectious effect
• No stopping rules replication of negative and
  positive findings will continue to be published
  for as long as there is interest

59
Association between p53 codon 72 polymorphism and
squamous cell cervical cancers
Koushik et al., CEBP 2004
60
Corroboration of Epidemiologic Findings

• The downside epidemics of repetition
• Genetic association studies have become more
  ambitious
• Early studies focused on one or a few candidate
  SNPs
• Recent studies target many SNPs and haplotypes
  using high throughput platforms
• Solution Bayesian approaches, e.g., false
  positive report probability (Wacholder et al.,
  JNCI 2004)
• FPRP Probability of no association given a
  statistically significant finding for a putative
  association
• Based on 3 quantities prior probability that the
  association is true, p value for the finding,
  power of the study

61
AR for some established causal relations in
cancer
100
20
50
10
Attributable Proportion
5
2
1.5
Prevalence of risk factor
Franco Harper, Vaccine 2005
62
Proportion of cancers attributed to different
factors
Factor Best estimate () Plausible Range ()
Tobacco 33 25 - 40
Diet 30 20 - 60
Infection viral, bacterial, parasitic 18 10 - 25
Reproductive factors and hormones 7 5 - 10
Ionizing radiation 6 4 - 8
Heredity 5 2 - 8
Occupation 3 2 - 8
Alcohol 3 2 - 4
UV light 1 0.5 - 1
Pollution lt1 lt1 - 2
Medicines lt1 lt1- 2
Industrial products lt1 lt1 - 2
Food additives lt1 -2 - 1
Sources Doll Peto, 1981 1996 Levine et al,
1989 Li et al., 1991 Pisani et al., 1997 Key
et al., 1997 Parkin et al., 2006
63
(No Transcript)
64
(No Transcript)
65
(No Transcript)
66
(No Transcript)
67
ORs of upper aero-digestive tract cancer in
southern Brazil according to joint exposure to
tobacco and alcohol consumption. Results by
conditional logistic regression (matching
variables age, sex, study location, and
admission period) controlling for race,
temperature of beverages, religion, use of a wood
stove, and consumption of spicy foods. Model A
assumes independence of effects. Model B assumes
effect modification. Levels of lifetime alcohol
consumption 1) lt1 2) 1-145 3) 146-932 4) gt932
kgs levels of cumulative tobacco exposure 1)
never smoked 2) 1-25 3) 26-60 4) gt60
pack-years. Source Schlecht et al., Am J
Epidemiol, 1999
68
VIRUSES IMPLICATED AS CAUSES OF HUMAN CANCER
Virus Group (genome) Convincingly linked to Possibly implicated in
Hepatitis B virus (HBV) Hepadnavirus (3 Kb DNA) Liver
Hepatitis C virus (HCV) Flavivirus (10 Kb RNA) Liver NH lymphoma (NHL), cryoglobulinemia, monoclonal gammopathy
Human papillomavirus (HPV) Papillomaviridae (8 Kb DNA) Cervix, anogenital Oral, skin
Simian virus 40 (SV 40) (also JC and BK viruses) Polyomaviridae (5 Kb DNA) Mesothelioma, CNS, osteosarcoma, NHL (SV40?)
Human T Lymphotropic viruses (HTLV) Retrovirus (10 Kb RNA) T-cell leukemias
Human immunodeficiency virus (HIV) Retrovirus (10 Kb RNA) AIDS-associated malignancies
Epstein-Barr virus (EBV, HHV-4) Gamma-herpesvirus (170 Kb DNA) NHL, nasopharynx Hodgkins lymphoma, breast
Herpes simplex virus 2 (HSV-2, HHV-2) Alpha-herpesvirus (150 Kb DNA) Cervix (cofactor?)
Cytomegalovirus (CMV, HHV-5) Beta-herpesvirus (230 Kb DNA) Cervix (cofactor?)
Human herpesvirus 8 (KSHV, HHV-8) Gamma-herpesvirus (140 Kb DNA) Kaposis sarcoma Castleman's disease, Pleural effusion lymphoma
Human herpesvirus 6 (HHV-6) Beta-herpesvirus (160 Kb DNA) NHL (?)
69
(No Transcript)
70
(No Transcript)
71
(No Transcript)
72
Criteria used in attributing causality to
candidate microbial agents
Evans (1976)
Antibody to the agent is regularly absent prior to the disease and exposure to the agent
Antibody to the agent regularly appears during illness and includes both immunoglobulins G and M
Presence of antibody to the agent predicts immunity to the disease associated with infection by the agent
Absence of antibody to the agent predicts susceptibility to both infection and the disease produced by the agent
Antibody to no other agent should be similarly associated with the disease unless a cofactor in its production
Evans and Mueller (1990)
Geographic distributions of viral infection and tumor should coincide
Presence of viral marker should be higher in cases than in controls
Incidence of tumor should be higher in those with the viral marker than in those without it
Appearance of viral marker should precede the tumor
Immunization with the virus should decrease the subsequent incidence of the tumor
Fredricks and Relman (1996)
Nucleic acid belonging to putative pathogen should be present in most cases and preferentially in organs known to be diseased
Few or no copy numbers should occur in hosts or tissues without disease
Copy number should decrease or become undetectable with disease regression (opposite with relapse or progression)
Detection of DNA sequence should predate disease
Microorganism inferred from the sequence should be consistent with the biological characteristics of that group of organisms
Tissue-sequence correlates should be sought at the cellular level using in situ hybridization
Above should be reproducible
Adapted from Franco et al., SCB 2004
73
Evaluation of Carcinogenicity to Humans IARC
Monograph Series
Infectious Agent Volume, year Evaluation Group
Hepatitis B Virus (HBV) (chronic infection) 59, 1994 Carcinogenic 1
Hepatitis C Virus (HCV) (chronic infection) 59, 1994 Carcinogenic 1
Hepatitis D Virus (HDV) 59, 1994 Not classifiable 3
Schistosoma haematobium 61, 1994 Carcinogenic 1
Opistorchis viverrini 61, 1994 Carcinogenic 1
Clonorchis sinensis 61, 1994 Probably carcinogenic 2A
Schistosoma japonicum 61, 1994 Possibly carcinogenic 2B
S. mansoni 61, 1994 Not classifiable 3
O. felineus 61, 1994 Not classifiable 3
Helicobacter pylori 61, 1994 Carcinogenic 1
Human papillomavirus (HPV) types 16 and 18 64, 1995 Carcinogenic 1
HPVs types 31 and 33 64, 1995 Probably carcinogenic 2A
HPVs, other types (except 6/11) 64, 1995 Possibly carcinogenic 2B
Human Immunodeficiency Virus (HIV) type 1 67, 1996 Carcinogenic 1
Human T Lymphotropic Virus (HTLV) type I 67, 1996 Carcinogenic 1
HTLV-II 67, 1996 Not classifiable 3
HIV-2 67, 1996 Possibly carcinogenic 2B
Epstein-Barr Virus (EBV) 70, 1997 Carcinogenic 1
Human Herpesvirus (HHV) type 8 70, 1997 Probably carcinogenic 2A
HPV 16, 18, 31, 33, 35, 39, 45, 51, 52, 56, 58, 59, 66 90, 2007 Carcinogenic 1
HPVs 6, 11 90, 2007 Possibly carcinogenic 2B
HPV genus Beta 90, 2007 Possibly carcinogenic 2B
74
Infection-related cancers worldwide estimates
for 2002 (Parkin, IJC 2006)
Agent Agent Cancer Number of cases of all cancers
Viruses HPV Cervix 492,800 5.2
Viruses HPV Ano-genital 53,880 5.2
Viruses HPV Oral 14,500 5.2
Viruses HBV and HCV Liver 535,000 4.9
Viruses EBV Nasopharynx 78,100 1.0
Viruses EBV Hodgkin lymphoma 28,600 1.0
Viruses EBV Burkitt's lymphoma 6,700 1.0
Viruses HIV / HHV-8 Kaposi's sarcoma 66,200 0.9
Viruses HIV / HHV-8 Non-Hodgkin's lymphoma 36,100 0.9
Viruses HTLV-I ATL 3,300 0.03
Bacteria H.pylori Stomach 592,000 5.5
Bacteria H.pylori Lymphoma 11,500 5.5
Helminths Schistosomes Bladder 10,600 0.1
Helminths Liver flukes Liver 2,500 0.02
All agents All agents All agents 1,932,800 17.8
75
Infection-related cancers worldwide estimates
for 2002 (Parkin, IJC 2006)
Cancer Agent Developed countries Developed countries Developed countries Developing countries Developing countries Developing countries
Cancer Agent Total cancers Agent-attributable cancers all cancer Total cancers Agent-attributable cancers all cancer
Liver HBV,HCV 110,400 48,000 1.0 513,100 475,000 8.2
Liver Flukes 0 2,500
Cervix HPV 83,400 83,400 1.7 409,400 409,400 7.0
Stomach H.pylori 311,200 192,000 3.8 619,200 400,000 6.9
KS HIV/HHV8 3,700 0.1 62,500 1.1
NHL H.pylori 151,100 5,600 0.2 149,200 5,900 0.7
NHL EBV 151,100 100 0.2 149,200 6,600 0.7
NHL HIV 151,100 9,300 0.2 149,200 26,800 0.7
NHL HTLV-I 151,100 550 0.2 149,200 2,790 0.7
Anogenital HPV 38,000 22,450 0.4 58,700 31,430 0.5
Nasopharynx EBV 7,200 6,500 0.1 72,600 71,600 1.2
Oral HPV 115,500 5,600 0.1 210,700 8,800 0.2
Hodgkin lymphoma EBV 28,000 11,500 0.2 34,300 17,100 0.3
Bladder Schistosomes 225,200 0 0.0 131,000 10,600 0.2
All cancers All agents 5,016,000 389,000 7.7 5,828,000 1,527,000 26.3
76
(No Transcript)
77
(No Transcript)
78
(No Transcript)
79
(No Transcript)
80
The squamo-columnar Junction and Transformation
Zone of the Uterine Cervix
81
Progression to cervical cancer
Years
Decades
Months
CIN1
Normal Epithelium
HPV Infection
CIN2
CIN3
Carcinoma
HSIL
ASCUS/LSIL
Screening
Treatment
SIL Squamous intraepithelial lesion
CIN Cervical intraepithelial neoplasia
82
(mucosal and cutaneous PVs of humans and primates)
Species A10 HPVs 6, 11 and related
Species A7 HPV 18 and related
Species A9 HPV 16 and related
(cutaneous PVs of humans)
(cutaneous PVs of humans)
De Villiers et al., Virology 2004
83
(No Transcript)
84
Relative Risk estimates from the pool of IARC
case-control studies Muñoz et al., NEJM
2003 Graph kindly provided by the Editors of
HPV Today
85
(No Transcript)
86
Summary of conclusions World Cancer Research
Fund / American Institute for Cancer Research.
Food, Nutrition, Physical Activity, and the
Prevention of Cancer a Global Perspective.
Washington DC AICR, 2007
87
(No Transcript)
88
(No Transcript)
89
(No Transcript)
90
(No Transcript)
91
(No Transcript)
92
(No Transcript)
93
(No Transcript)
94
(No Transcript)
95
(No Transcript)
96
(No Transcript)
97
(No Transcript)
98
(No Transcript)
99
Efficacy versus Effectiveness

• Efficacy
• Assessment of screening strategy under ideal
  conditions of test performance in controlled,
  investigational settings
• Effectiveness
• Assessment of screening strategy in actual public
  health conditions that reproduce the complete
  context of test deployment and post detection
  intervention

100
(No Transcript)
101
(No Transcript)
102

• BIASES IN SCREENING
• Selection bias (all designs)
• Referral (volunteer) bias, length-biased sampling
• Lead time bias (all designs)
• Overdiagnosis bias (all designs)
• Verification bias (all designs)
• False gain in sensitivity due to test combination
  (Franco, 2000)
• Sticky diagnosis and slippery linkage biases
  (RCTs) (Black et al., 2002)

103
(No Transcript)
104
Length-biased sampling
105
Lead time bias
Adapted from Mittra, 1993
106
(No Transcript)
107
Bias due to differential verification based on
screening results
Table of screening results if only a sample is
tested 80 for test and 10 for test-
Franco, Lab Clin N Amer, 2000
108
Verification bias
Franco, Lab Clin N Amer, 2000
109
(No Transcript)
110
Franco Ferenczy, AJOG 1999
111
Franco Ferenczy, AJOG 1999
112
Interpreting gain in sensitivity and loss in
specificity when HPV testing is added to Pap
cytology in cervical lesion triage
Diagnostic utility ()
Significance versus
PapHPV (95CI)
expected Pap chance
Pap alone
expected Pap chance
Study
Index
Pap alone
Cox 1992
Sensitivity
44
78 (71-84)
60
yes ()
yes ()
Specificity
92
79 (75-83)
65
yes (-)
yes ()
Hatch 1995
Sensitivity
76
92 (86-95)
89
yes ()
no
Specificity
57
43 (36-51)
27
yes (-)
yes ()
Franco Ferenczy, AJOG 1999
113
Sticky Diagnosis Bias(Black et al., JNCI 2002)

• In an RCT, the target cancer is more likely to be
  detected in the screened group than in the
  control group
• Deaths are more likely to be attributed to the
  target cancer in the screened group
• Example Excess lung cancer mortality in the
  screened arm of the Mayo Lung Project

114
Slippery Linkage Bias(Black et al., JNCI 2002)

• In an RCT, more subjects undergo invasive
  procedures and treatment in the screened group
  than in the control group
• These interventions may lead to deaths which may
  not be assigned to the screening intervention
  (i.e., they slip away from appropriate linkage)
• Example Excess cardiovascular deaths in the
  screening arm of the Minnesota Colon Cancer Study

115
Measures and surrogates of improved outcome for
determining screening efficacy and effectiveness

1. Decrease in cause-specific mortality
2. Reduction in incidence of advanced cancers
3. Increase in survival
4. Shift in stage to early cancers
5. Enhanced detection of precursor lesions

116
(No Transcript)
117
(No Transcript)
118
Decrease in death rate (per 100,000 women)
between 1960-62 and 1970-72
Average screening rate (per 1000 women)
Relationship between intensity of Pap cytology
screening and decrease in mortality from cervical
cancer in Canadian provinces (Source Boyes et
al., 1977 WHO)
119
(No Transcript)
120
RCTs on mammography screening among women aged
40-74
Trial Allocation No. women Accrual Age at entry Method Screening interval Attendance rate
HIP, USA Individual 60 995 1963-96 40-64 M,CBE 12 ms 67
Malmo I, Sweden Individual 42 283 1976-93 45-70 M 18-24 ms 74
Malmo II, Sweden Individual 17 793 1963-96 43-49 M 18-24 ms 75-80
2-county trial, Sweden Cluster, district 56 448 1978-94 40-74 M 24,33 ms 89
Oestergotland Cluster, parish 76 617 1977-86 40-74 M 24-33 ms 89
Edinburgh, Scotland Cluster, GP practices 44 268 1978-88 45-64 M,CBE 24 ms 61
NBSS I, Canada Individual 50 430 1978-85 40-49 M,CBE,P 12 ms 100
NBSS II, Canada Individual 39 405 1980-85 50-59 M,CBE,P 12 ms 100
Stockholm, Sweden Cluster, DOB 60 117 1981-86 40-64 M 28 ms 81
Goteborg, Sweden Individual 51 611 1982-91 39-59 M 18 ms 84
Finland Cluster, DOB 158 755 1987-90 50-64 M 24 ms 90
IARC, 2002
121
Efficacy of screening for breast cancer by
mammography alone in women aged 40-49
Trial invitations to screening RR of death (95 CI)
Malmo I, Sweden 4 in 8 yrs 0.74 (0.44-1.25)
Malmo II, Sweden 4 in 8 yrs 0.65 (0.39-1.08)
Kopparberg, Sweden 3 in 6 yrs 0.76 (0.42-1.40)
2-county trial, Sweden 4 in 8 yrs 1.05 (0.64-1.71)
Stockholm, Sweden 2 in 4 yrs 1.52 (0.80-2.88)
Goteborg, Sweden 3 in 5 yrs 0.58 (0.35-0.96)
All trials 0.81 (0.65-1.01)
IARC, 2002
122
Efficacy of screening for breast cancer by
mammography alone in women aged 50-69
Trial invitations to screening RR of death (95 CI)
Malmo I, Sweden 4 in 8 yrs 0.84 (0.68-1.04)
Kopparberg, Sweden 3 in 6 yrs 0.52 (0.39-0.70)
2-county trial, Sweden 4 in 8 yrs 0.81 (0.64-1.03)
Stockholm, Sweden 2 in 4 yrs 0.68 (0.44-1.04)
Goteborg, Sweden 3 in 5 yrs 0.94 (0.62-1.43)
Finland 2 in 4 yrs 0.76 (0.53-1.09)
All trials 0.75 (0.67-0.85)
IARC, 2002
123
(No Transcript)
124
5-year relative survival for all sites of cancer,
children versus all ages, US SEER program.
Source Ries et al., 2005
125
(No Transcript)
126
(No Transcript)
127
Using Mediated Analysis to Assess Prognostic
Pathways

• Hypotheses
• Race/ethnicity associated with stage and
  treatment
• Association between race/ethnicity and survival
  is indirect via the prognostic effects of stage
  and treatment

128

• Hypotheses
• Race/ethnicity associated with stage and
  treatment
• Association between race/ethnicity and survival
  is indirect via the prognostic effects of stage
  and treatment

Franco et al., J. Clin. Epidemiol. 1993
129

• Observational study in a population-based
  clinical breast cancer register of one
  health-care region in Sweden
• N 9,059 women aged 5084 years diagnosed with
  primary breast cancer between 1992 and 2002
• 5-y relative survival ratio estimated by age
  group, diagnostic activity, tumor
  characteristics, and treatment

130
Prognostic model that explains the lower breast
cancer survival in women older than 70 years
Remote, distal prognostic effect of age is
mediated via three intermediate, proximal
determinants of clinical outcome. Ideally, only
co-morbidity and other age-related factors that
are beyond the reach of modern health care should
be operative. Unfortunately, decreased access to
screening and to optimal treatment and management
tend to further increase the disparity in
survival between young and old women with breast
cancer.
Franco, PLoS-Med 2006
131
5-year excess mortality rate (95CI) for breast
cancer patients aged 70-84 years relative to
those 50-69 years (Eaker et al., PLoS-Med 2006)
Model (adjustment) Stage IIB Stage III Unstaged
Year of diagnosis only 1.43 (1.091.87) 1.60 (1.132.26) 1.83 (1.272.63)
Year of diagnosis, lymph node involvement, tumor size, and treatment 1.26 (0.891.80) 0.91 (0.561.48) 1.07 (0.731.56)
132
Additional Slides

• Supplement to points discussed in the articles

133
Using Mediated Analysis to Assess Etiologic
Pathways
JNCI 85 958-964, 1993
Sexual behaviour
HPV infection
CIN
From table 2 RR1 adjusted for age in
sextiles RR2 adjusted for age, age at 1st
intercourse, education, income, smoking, OC use,
parity RR3 adjusted for age and HPV test results
134
Gorey KM, Holowary EJ, Fehringer G, Laukkanen E,
Moskowitz A, Webster DJ, Richter NL. An
international comparison of cancer survival
Toronto, Ontario, and Detroit, Michigan,
metropolitan areas. Am J Public Health. 1997
Jul87(7)1156-63.
Lack of access to screening and early diagnosis
More advanced stage at diagnosis
Poor prognosis and survival
Low SES
Lack of access to best treatment
Hypothesis SES has a differential effect on the
survival of adults diagnosed with cancer in
Canada and the United States Ontario Cancer
Registry and US NCI's SEER program provided a
total of 58,202 and 76,055 population-based
primary cancer cases for Toronto and Detroit,
respectively SES data for each person's residence
taken from population censuses Compared 1- and
5-yr survival rates by low, middle, and high SES
(contextual)
135
Cervical Cancer Survival Association with
Socioeconomic Status
5-year survival rate ()
Income level
From Gorey et al., AJPH, 1997
136
In the US cohort, there was a significant
association between SES and survival for 12 of
the 15 most common cancer sites (low
SESworse). In the Canadian cohort, only 3 of the
15 sites showed an association but with no clear
trend. Patients of low-income areas in Toronto
experienced a survival advantage for 13 of 15
cancer sites at 1- and 5-year follow-up.
Gorey et al., AJPH 1997
137
Selective Reporting Biases in Cancer Prognostic
Factor Studies Panayiotis A. Kyzas , Konstantinos
T. Loizou , John P. A. Ioannidis Journal of the
National Cancer Institute 2005 971043-1055

• Identified all TP53 and HNSCC papers published
  until April 2004 with and without indexing on
  mortality or survival terms
• Used meta-analysis techniques to identify
  potential biases in prognostic effects
  attributable to TP53 positivity (IHC or molecular
  methods)

138
Selective Reporting Biases in Cancer Prognostic
Factor Studies Panayiotis A. Kyzas , Konstantinos
T. Loizou , John P. A. Ioannidis Journal of the
National Cancer Institute, 97 (14, July 20,
2005) 1043-1055
116 studies with TP53 status in HNSCC
20 excluded overlapping
17 excluded no clinical data
79 eligible with clinical data
15 mortality not alluded
64 clinical data and mortality alluded, authors
contacted
22 data not available
42 with evaluable prognostic analysis
18 published, survival indexed
13 published, survival not indexed
11 data retrieved and analyzed
139
Selective Reporting Biases in Cancer Prognostic
Factor Studies Prognostic Role of TP53 status in
HNSCC Kyzas et al. Journal of the National Cancer
Institute, 2005 97 1043-1055
Study set Number of studies RR of death (95CI)
All 42 1.16 (0.99-1.35)
All published 31 1.23 (1.03-1.47)
Published and indexed 18 1.27 (1.06-1.53)
Published, not indexed 13 1.13 (0.81-1.59)
Data retrieved 11 0.97 (0.72-1.29)
140
Selective Reporting Biases in Cancer Prognostic
Factor Studies Panayiotis A. Kyzas , Konstantinos
T. Loizou , John P. A. Ioannidis Journal of the
National Cancer Institute, 97 (14, July 20,
2005) 1043-1055
Study feature Number of studies RR of death (95CI)
Blinding used 20 1.05 (0.86-1.28)
Blinding not stated 22 1.32 (1.06-1.65)
Prospective design 6 1.01 (0.71-1.43)
Retrospective design 31 1.22 (1.00-1.49)
Design unclear 5 1.18 (0.91-1.53)