Reducing the Burden of TobaccoAssociated Cancers in Florida

1
Reducing the Burden of Tobacco-Associated Cancers
in Florida Florida Cancer Data System,
Sylvester Comprehensive Cancer Center Department
of Epidemiology and Public Health Department of
Surgery, Division of Surgical Oncology Florida
Department of Health, Bureau of Epidemiology
2
(No Transcript)
3

• Team Science Program Grant Program
• Awarded to multidisciplinary, multi-investigator
  and multi-project programs at larger research
  institutions
• Goal is to assist them in building a research
  infrastructure
• One million dollars over a 24 month period

4

• 2006-7 Funded Team Science Programs
• University of Florida- (PI Paul Davenport) The
  Role of Nicotine in the Neural Control of
  Respiratory and Cardiovascular Systems (949,404)
  
• Mayo Clinic- (PI Alan Fields) Oncogenic PKC iota
  in Smoking-Related Lung Cancer , 950,000
• University of Miami- (PI David Lee) Reducing the
  Burden of Tobacco-Associated Cancers in Florida ,
  949,974

5

• Project Aims
• Identify areas of Florida in greatest need of
  primary prevention by mapping county-level
  smoking prevalence data
• Identify areas of Florida in greatest need of
  secondary and tertiary prevention by using
  spatial analysis methods that will illustrate
  tobacco-associated cancer clusters
• Identify areas in Florida in need of secondary
  and tertiary prevention by using data from FCDS
  and from the Florida Agency for Health Care
  Administration to map late-stage presentation of
  tobacco-associated cancers and to evaluate access
  to state-of-the-art medical care for
  tobacco-related cancers.

6
(No Transcript)
7

• Outreach, Education and Dissemination Core
• (Dr. Lee PI Ms. Parker, Co-PI Dr. Kobetz Co-I)
  
• Will be responsible for laying the groundwork for
  communication of the knowledge acquired in
  Projects 1, 2 and 3 to the professional and lay
  community via the development of a website.
• Will also be instrumental in bringing together
  key stakeholders and community leaders into the
  Year 2 activities leading up to the development
  of the NCI Program Project application.
• Will also begin the process of identifying
  community assets and barriers in preparation for
  the P01 intervention project.

8

• Biostatistics and Data Management Core
• (Dr. Anello PI Dr. MacKinnon, Co-PI)
• Provides data management and statistical analytic
  support for the three projects including the
  development of a web-based query system for the
  Website.
• Will work in consultation with Dr. Kuldorff, a
  spatial statistician and creator of the
  time/space scan statistical package (SaTScan), to
  apply this methodology to identify areas in
  Florida in need of primary, secondary, and
  tertiary tobacco-associated cancer prevention.
  Will also examine other small area estimation
  techniques.

9

• Project 1
• (Dr. Huang PI Dr. Voti Co-PI Dr. Dietz Co-I)
• Collect data on tobacco use related behaviors
  among middle high school students in Florida as
  part of the 2006 FYTS. Conduct in 2007 a combined
  BRFSS and FLATS survey with enhanced sample size.
• Explore various statistical methodologies to
  analyze the adult and youth tobacco data,
  estimate the prevalence of youth and adult
  tobacco use related behaviors in sub-groups of
  interest (e.g., gender, race-ethnicity) at the
  county-level and below.

10

• Project 1
• (Dr. Huang PI Dr. Voti Co-PI Dr. Dietz Co-I)
• Provide area- and county-specific data to Project
  2 and to the Outreach, Education, and
  Dissemination Core to inform the design of
  primary prevention interventions for the future
  NCI Program Project.

11
2002 County-Level Smoking Prevalence Quartiles
12

• Project 2
• (Dr. Fleming PI Dr. Mackinnon Co-PI)
• To utilize the cancer cluster software program
  SaTScan to analyze Floridas tobacco-related
  cancer incidence and mortality data identify
  geographic areas (i.e. census block groups) with
  higher than expected tobacco-related cancer
  rates
• Using US Census SES indicators and Project 1
  smoking rates, assign an SES category and a
  smoking prevalence rate category to each block
  group in Florida for analyses designed to examine
  factors related to excess cancer incidence and
  mortality rates.

13

• Project 2
• (Dr. Fleming PI Dr. Mackinnon Co-PI)
• Provide state incidence and mortality information
  to Project 3, and state incidence and mortality
  maps to the Outreach, Education, and
  Dissemination Core on all subgroups of interest
  (e.g., age, gender, race/ethnic, etc.) to inform
  to the design of the future NCI Program Project.

14

• SatScan
• SaTScan employs a spatial scan statistic which
  can be used to identify the location of cancer
  clusters
• Modeling approach permits adjustment for
  covariates and interaction terms

15
(No Transcript)
16

• Project 3
• (Dr. Franceschi PI Dr. Koniaris Co-PI Dr.
  McCollister, Co-I)
• Use SaTScan to identify regions within Florida
  which have a higher than expected incidence of
  late stage tobacco-associated cancer.
• Utilizing Agency for Health Care Administration
  (ACHA) data sets, identify patient- and
  treatment-level factors associated with outcome
  for tobacco associated cancers.

17

• In collaboration with the Outreach Education and
  Dissemination Core, create the framework to
  provide educational opportunities directed at
  primary and secondary health care providers in
  areas of need to make them aware of problems of
  late diagnosis, and options to improve early
  detection and cancer treatment outcomes.

18

• Treatment Center Volume Improves Outcomes for
  Pancreatic Cancer Care
• Group I treated gt 30 patients/year (n 9
  centers)
• Group II treated between 10 and 30 patients/year
  (n 61 centers)
• Group III treated less than 10 patients per year
  (n 177 centers)

19
Overall Survival Surgical Patients

20

• Long-Term Objectives
• Draw upon expertise to reduce tobacco-associated
  cancer incidence and mortality in high risk
  regions in Florida.
• This effort will be organized around the three
  levels of prevention
• Primary
• Secondary
• Tertiary
• Seek funding from the NCI/ACS to pursue these
  studies via a series of R01s and/or thru a
  P01-type mechanism

21

• Our Roadmap
• Quarterly meetings with the IAC also quarterly
  conference calls with the EAC plus an EAC annual
  meeting (first is scheduled for the end of
  October).
• Development of intervention-focused R01s or P01
  for submission in October, 2007 (fallback
  February, 2008?).

22
Thank You
23
Biostatistics and Data Management Core (Co-PI
C. Anello and J. MacKinnon)
24

• Biostatistics Core (BC) Outline
• Level of Support
• Purpose of BC
• Action Plan

25

• Level of Support
• PI is 10 effort
• 1 Biostatistician FTE at 10 effort
• 1 Statistical research associate at 20 effort.
  No cost

26

• Purpose of BC
• Provide statistical expertise to support the aims
  of the 3 identified projects
• Work with (Youjie Huang and Lydia Voti) on
  methodological issues related to BRFSS, FLATS,
  and FYTS surveys analysis, as needed
• Work with Drs Fleming and MacKinnon on
  alternative methods to the current spatial
  statistical methodology, allowing for
  incorporation of other data
• Work with Drs. Franceschi and Koniaris on
  comparing treatment and outcomes in relation to
  geographic regions and medical facility volume.
  Taking into account potential confounders such as
  co-morbidity

27

• Action Plan
• Work on statistical aspects of small area
  estimation Using data from 2000 BRFSS survey to
  investigate various approaches to estimating
  county level data.
• Examine the role of hierarchical modeling as an
  analyses tool in combining information from
  various sources.
• Examine assumptions of the various models used in
  the analyses in relation to available data
• Cluster analysis- SaTScan/Dr. Kulldorff
• Small sample estimation
• Survival analyses and long rank tests

28
Action Plan (cont.) 4. Review analysis and
validity, when possible 5. Review inferences
from analysis
29

• Some Special Statistical Issues
• Comparison of small area estimations methodology
• Role of hierarchical modeling
• Analysis of outcome in relation to medical
  facility volume taking into account potential
  confounders such as co-morbidity

30
Data Management

• Assemble, merge, manipulate, and make various
  databases available to Team Science Researchers

31
Data Bases

• FCDS Commercial File
• Tobacco related sites
• AHCA
• Merged with FCDS data
• Assignment of co-morbid conditions
• BRFSS
• FLATS
• FYTS

Raw and Aggregated Data
32
Data Access

• Team Science web site
• Public access
• Tabular data
• Maps
• Publications
• Password protected access
• Team Science research team members only

33
Thank You
34
Outreach, Education and Dissemination
Core (OEDC) David Lee, PI Dorothy Parker,
Co-PI Erin Kobetz, Co-I Communication with
professional and lay audiences to promote and
develop tobacco-control research aimed at
reducing morbidity and mortality.
35
OEDC Goals

• Disseminate findings from Projects 1, 2 and 3 to
  researchers, health care providers, and the
  general public.
• Help develop future studies in areas identified
  as high risk with input from key stakeholders and
  community leaders in those areas.

36
Research Health Care Audience

• FTRG researchers
• Access to raw data for analyses (internal web
  page)
• Other researchers internal external
• Build interdisciplinary research team
• Medical public health professionals
• Increasing awareness of high risk areas
• Data for planning and evaluating interventions
• 1, 2 3 prevention
• Statewide cancer control groups programs
• CCRAB, FCC, FDOC, Regional Collaboratives, etc.

37
Public/Lay Audiences

• Advocacy groups and CBOs interested in tobacco
  issues
• Local community groups and organizations in high
  risk areas interested in education and cessation
• among all age groups
• The media (TV, radio, newspapers)
• State and local government
• elected officials

38
http//tobaccoinfo.med.miami.edu
39
Input for Future Research

• Feedback on web page
• Ongoing changes to make it useful and
  user-friendly for public and researchers
• Input from high risk communities
• Public forums, focus groups, on-line surveys
• Appropriateness and acceptability of proposed
  interventions

40
Role in Future Research

• Population Research Core and other shared
  resources at SCCC
• Disparities
• Multicultural communities
• Community-based participatory research

41
Thank You
42
Project 1 (Dr. Huang PI Dr. Voti Co-PI Dr.
Dietz Co-I)
43

• Project 1 Objectives
• Complete in 2007 a county-level Behavioral Risk
  Factor Surveillance System (BRFSS) survey with
  expanded tobacco-related questions.
• Collect data on tobacco use related behaviors
  among middle school and high school students in
  Florida as part of the 2006 FYTS.
• Analyze data from the BRFSS, FLATS and FYTS to
  identify areas with high rates of tobacco use.

44
Background BRFSS

• A population-based, random telephone survey,
  following CDC survey protocol.
• Sample stratified by size of county population
  and over-sampled for minorities
• Monitors behavioral risk factors and chronic
  diseases among residents 18 years and older.
• Conducted annually since 1986.
• Sample size was approximately 9,000 in 2006
• Supported by CDC.

45
Background BRFSS Questions

• Health Status
• Health-Related Quality of Life
• Health Care Access
• Demographics
• Exercise
• Diabetes
• Hypertension Awareness
• Cholesterol Awareness
• Cardiovascular Disease
• Asthma
• Immunization

• Tobacco Use
• Alcohol Consumption
• Disability
• Arthritis Burden
• Fruits Vegetables
• Diabetes
• Actions to Control High Blood Pressure
• Womens Health
• Prostate Cancer Screening
• Colorectal Cancer Screening

46
Background Adult Tobacco Survey (ATS)

• A random telephone survey of adults for
  monitoring smoking and tobacco-related health
  behaviors
• Sponsored by the Centers for Disease Control and
  Prevention (CDC)
• Has been administered annually in Florida since
  2003
• More than 100 questions
• Sample size 2,600-4,000

47
Background ATS Questions

• Demographics
• Tobacco use cigarettes, smokeless tobacco, and
  cigars
• Cessation
• Exposure to second-hand smoke
• Risk Perception and social Influences
• Health and social influences
• State-added questions

48
Plan of BRFSS Survey

• To increase sample size to 26,000 for estimating
  county level prevalence of smoking
• To increase sample size by 6,000 using the grant
  funding in 2007
• To combine BRFSS and ATS survey by adding 4,000
  sample size in 2007
• To combine 2006 BRFSS data that adds 9,000 sample
  size

49
Progress of BRFSS Survey

• DOH plans a county-level BRFSS survey in 2007
• Total sample size gt38,5000
• At least 500 completed interviews per county
• Approximately 150 questions
• 5 counties added county-specific question
• 11 counties increased sample size
• The questionnaire is finalized
• Currently working with CDC and survey company to
  implement the plan
• Data will be available in April 2008

50
Progress of Adult Tobacco Survey

• To best use the resource from the grant
• Plan A To increase sample size and over sample
  Hispanic population
• To obtain unique data among Hispanics
• Plan B To increase sample size and over sample
  areas with high prevalence of smoking and/or
  smoking attributable cancer
• To study areas at greater risk
• Plan C To conduct a follow-up survey using the
  County BRFSS information
• Cost effective

51
Background FYTS

• A self-administered, confidential, school-based
  random survey of Florida public middle and high
  school students
• Monitors tobacco use behaviors, knowledge and
  attitude, as well as exposure to second hand
  smoking
• Approximately 100 questions
• Conducted annually in March-April since 1998
• A multi-agency collaboration

52
Background FYTS Questions

• Tobacco use (cigarette, smokeless tobacco, cigar,
  bidis, and kreteks)
• Tobacco use prevention education in school
• Students knowledge about and attitudes toward
  tobacco use
• Influence of anti-tobacco media and from tobacco
  companies
• Students exposure to second-hand smoke

53
Background 2002 BRFSS

• County-level BRFSS
• Random sample surveys by county
• Conducted September 2002 to January 2003
• 500 completed interviews per county (Total
  34,551)
• Core questions 76 questions
• State-level BRFSS
• A random statewide survey
• Implemented from 1/1/2002 to 12/31/2002
• Core questions, optional module and state added
  questions
• Sample size 6,150

54
Progress of FYTS Survey

• 2006 survey was completed in May 2006
• 783 schools in 65 counties participated in the
  survey in 2006
• Response rate
• 73 in middle schools
• 68 in high schools
• Sample size in 2006 96,000 in 65 counties
• 33,112 middle school students
• 29325 high school students

55
Plan of Analysis of Data

• Explore various statistical methodologies to
  analyze the combined BRFSS and ATS adult survey
  data and the 2006 FYTS data, estimate the
  prevalence of youth and adult tobacco use related
  behaviors in sub-groups of interest (e.g.,
  gender, race-ethnicity) at the county-level and
  possibly below the unit of County.

56
Progress of Data Analysis

• FYTS data
• Data has been cleaned and weighted by CDC.
• Data were received in September
• Data were analyzed at state-level and county
  level
• A state-level report and 65 county reports are
  under DOH review
• The data are available for research

57
Progress of Data Analysis

• Adult data (BRFSS and ATS)
• Gathering information and data for analysis
  completed
• Developing methodologies for estimating
  prevalence at small areas using 2002 BRFSS
  (County BRFSS and State BRFSS)
• Synthetic estimate
• Direct estimate with combined sample
• Models that including Census data
• Analyzing 2007 data

58
Thank You
59
Project 2 Co-Principal Investigators Dr. Jill
A. MacKinnon and Dr. Lora E. Fleming
60

• Special THANK YOU to Recinda Sherman for mapping
  the data

61
Goal
Using FCDS incidence and mortality data,
identify and characterize areas with higher than
expected burden for the following tobacco
associated cancers lung, oral, esophageal,
pancreas, larynx, cervical, kidney, bladder,
acute myeloid leukemia and stomach cancer.
62
Geographical Information Systems

• Using spatial statistical methodology
• Identify areas of higher than expected cancer
  incidence and mortality
• SaTScan
• Mapping county-level smoking prevalence
  (BRFSS/FLATS/FYTS)

63
Spatial Analysis - SaTScan

• Developed under the joint auspices of Dr. Martin
  Kulldorff, the National Cancer Institute and Dr.
  Farzad Mostashari at the New Your City Department
  of Health and Mental Hygiene.
• Spatial scan statistic
• Cluster detection test
• Detect location of clusters
• Evaluate their statistical significance

Consultant to the Team Science project
64
SaTScan Process

• County to Census block group level
• Age and sex covariates
• Files needed
• Cases
• Population
• Geographic centroid (county, census tract, block
  group)

65
SaTScan

• Monte Carlo techniques
• Assigns O/E probabilities to defined block groups
  
• Generates a number of random replications of the
  data set under the null hypothesis
• Test statistic is calculated for each random
  replication as well as for the real data set
• If the real data set is among the 5 percent
  highest, then it is identified as a cluster

66
SaTScan (cont)

• Poisson probability model
• 999 to 9,999 Monte Carlo replications (10
  population at risk)
• Expected n of cases
• Indirect standardization (State)
• Adjusts for covariates and interaction terms (age
  and sex)
• Spatial analysis only
• Identified areas with higher than expected number
  of cases

67
ID Specific Locations for Clinical/Prevention
Interventions

• Incidence
• Late stage presentation
• Mortality
• Overlay smoking and other risk factors
• Identification of sub-populations
• Outreach and education

68
Cancer Incidence Mapping
69
Florida Residents
Lung Cancer Incidence
Areas in red are statistically significant
clusters of block groups with higher than
expectedLung cancer incidence, P lt .10,
adjusted by age, Florida Residents diagnosed
1998-2002 999 Monte Carlo Replications
70
Invasive lung cancer incidence, Florida
1998-2002 999 Monte Carlo Repetitions, 10
percent population at risk, P value .05
71
Florida Residents
Lung Cancer Incidence
Areas in red are statistically significant
clusters of block groups with higher than
expectedLung cancer incidence, P lt .10,
adjusted by age, Florida Residents diagnosed
1998-2002 999 Monte Carlo Replications
72
Big cluster lung cancer incidence, Florida
1998-2002 999 Monte Carlo Repetitions, 10
percent population at risk, P value .05
73
Larynx Cancer Incidence
Oral Cancer Incidence
Esophageal Cancer Incidence
Cancer Incidence, 1998-2002 50 population at
Risk, P value .10
74
Oral Late Stage
Esophageal Late Stage
Cancer incidence, 50 population at risk, P lt
.05
75
Cancer Mortality
FCDS Passive Mortality Block Group
76
Lung Cancer FCDS Mortality, Florida 1998-2002 999
Monte Carlo Repetitions, 10 of population at
risk, P value .05
77
Laryngeal Cancer FCDS Mortality, Florida
1998-2002 999 Monte Carlo Repetitions, 50 of
population at risk, P value .10
78
Cancer Mortality
Vital Statistic Mortality County
79
Esophageal
Larynx
Lung
Oral
Cancer mortality, 1998-2002 9,999 Monte Carlo
repetitions, 50 population at Risk, P value .10
80
Preliminary Analysis
81
Thank You
82

• Project III
• (Dr. Dido Franceschi PI Dr. Leonidas Koniaris
  Co-PI Dr. McCollister, Co-I)
• Identify areas within Florida which are
  experiencing high rates of late stage
  presentation of tobacco-associated cancers.
• Identify and compare predictors of
  tobacco-associated cancer outcomes.

83

• Project Aims and Hypothesis
• Use SaTScan to identify regions within Florida
  which have a higher than expected incidence of
  late stage tobacco-associated cancer.
• Utilizing FCDS and Agency for Health Care
  Administration (ACHA) data sets, identify
  patient- and treatment-level factors associated
  with outcome for tobacco associated cancers.
• Controlling for patient-level and other
  characteristics, treatment at high volume
  institutions will lead to higher 12- month
  survival rates than treatment at low volume
  institutions.
• Receipt of palliative care will lead higher
  12-month survival rates, even after controlling
  for patient-level and other characteristics.

84

• Project Aims and Hypothesis (cont)
• In collaboration with the Outreach Education and
  Dissemination Core, create the framework to
  provide educational opportunities directed at
  primary and secondary health care providers in
  areas of need to make them aware of problems of
  late diagnosis, and options to improve early
  detection and cancer treatment outcomes.

85

• Treatment Center Volume Improves Outcomes for
  Pancreatic Cancer Care
• Evaluation of 18,966 patients from 1991-2000
• Group I treated gt 30 patients/year (n 9
  centers)
• Group II treated between 10 and 30 patients/year
  (n 61 centers)
• Group III treated less than 10 patients per year
  (n 177 centers

86
Pancreatic Cancer Care 18,996 Patients

87
Pancreatic Cancer Care 18,996 Patients
88
Primary Payer at Diagnosis ()
89
SEER Stage at presentation
90
Survival All Patients
Survival Surgical Patients
91
Overall - Survival
Surgical Patients - Survival
92
(No Transcript)
93

• In collaboration with project II use SaTScan to
  identify regions within Florida which have a
  higher than expected incidence of late stage
  tobacco-associated cancer.
• ACHA Linkage
• Started working with pancreas and esophageal
  dataset (1998 2005)
• Validate volume effect correcting for co-morbidity

94
Florida Females
Florida Residents
Florida Males
Esophageal Cancer Incidence
Areas in red are statistically significant
clusters of block groups with higher than
expectedEsophageal cancer incidence, P lt .10,
adjusted by age, Florida Residents diagnosed
1998-2002 999 Monte Carlo Replications
95
Florida Females
Florida Residents
Florida Males
Late Stage Esophageal Cancer
Areas in red are statistically significant
clusters of block groups with higher than
expectedlate stage Esophageal cancer incidence,
P lt .10, adjusted by age, Florida Residents
diagnosed 1998-2002 999 Monte Carlo Replications
96
421,269 entries for diagnosis codes and 120,764
procedure codes
ACHA Pancreas Dataset
269,098 unique diagnosis codes and 90,914 unique
procedure codes for individual patients
Group by Charlson or Elixhauser comorbidity
coding algorithms
97
Enhanced Charlson Comorbidity Distribution -
Overall
98
Unique Comorbidity coded for Unique Patient
99
Enhanced Charlson Comorbidity Distribution
Unique/patient
100
Enhanced Elixhauser Comorbidity Distribution
Unique/patient
101
(No Transcript)
102
Cox Regression
103
Thank You
104

• Our Roadmap
• Quarterly meetings with the IAC also quarterly
  conference calls with the EAC plus an EAC annual
  meeting.
• Development of intervention-focused R01s or P01
  for submission in October, 2007 (fallback
  February, 2008?).

105
Summary of August 31st IAC Meeting

• Use tobacco tax receipts as another way to
  characterize tobacco use throughout the State
• Explore utilizing available census bureau
  geographic data
• Consider breast cancer as tobacco-associated
  cancer
• Identify all tobacco-related projects in the
  University and take steps to form broader
  collaborative teams

106
Summary of August 31st IAC Meeting

• There was an extended discussion on the relative
  merits of seeking a P01 versus first seeking
  smaller grant mechanisms such as R21s and R01s.
• -One concern with first seeking a P01 is the lack
  of NCI-funded R01s among Project PIs.
• -Also mentioned was the current extreme
  competitive environment at NCI (driven by flat
  budgets in recent years).

107
Thank You
108
BRFSS
109
Esophageal Cancer Mortality, 1998-2002 9,999
Monte Carlo repetitions, P value .10
110
Laryngeal Cancer Mortality, 1998-2002 9,999 Monte
Carlo repetitions, P value .10
111
Lung Cancer Mortality, 1998-2002 9,999 Monte
Carlo repetitions, P value .05
112
Oral Cavity Cancer Mortality, 1998-2002 9,999
Monte Carlo repetitions, P value .10