Epidemiologic Studies of Cancer, Diabetes & Its
Treatments:Opportunities from Canada
Jeffrey A. JohnsonUniversity of Alberta
Edmonton, Canada
Cancer & Diabetes
EpidemiologyConsortium
June 14, 2010
Canadian Team
University of Alberta, Edmonton
Jeff Johnson Samantha Bowker Yutaka Yasui
University of British Columbia, Vancouver, BC
Carlo Marra
Welcome to Canada!~ 10 million km2 / ~3.9 million mi2
population: ~34 million (2009)
10 provinces, 3 territories
Health care is a national philosophy, but a P/T mandate
13 ‘health care systems’
All residents have publically funded insurance for ‘medically necessary’ health services (Canada Health Act)
Physician and Hospital services
Drug coverage varies (products, policy, population)
Providers send ‘bill’ to Health Ministry
Each resident has Personal Health Number
Health Care & DataCanadian ContextPublically Funded, Privately Delivered
Prescriber
Amt Dispensed
DIN
Unique ID
Mortality
Demographic Data
Hospital PhysicianInsurance Registry
Age
Gender
Location of Residence
Status Aboriginal
Co-Morbidities/Procedures
DM StatusIncident/Prev
DM StatusIncident/Prev
Co-Morbidities/Procedures
Drug
ICD-9-CMor
ICD-10CA
ICD-9-CMVital StatisticsCOD
Year of Database Initiation
Province Hospital Physician Drugs
Alberta 1973 1973 1994 (seniors)
BC 1985 1985 1997 (All)
Manitoba 1970 1970 1994 (All)
New Brunswick 1973 1989 1990 (Seniors)
Nova Scotia 1973 1991 1975 (Seniors)
Ontario 1963 1989 1994 (Seniors)
Prince Edward Island 1984 1989 --Quebec 1983 1986 1981 (Seniors)
Saskatchewan 1970 1971 1975 (All)
Case Definition of DMwith Large Admin
Databases
Case/Cohort Ascertainment Identifying diabetes in admin data
A. National Diabetes Surveillance System Case Def’n
2 physician visits for DM (ICD-9 250) in 2 year periodor
1 hospitalization for DM (ICD-10CA E10-E14)
B. Antidiabetic Drug Use
- e.g., new users of oral antidiabetic agents
0.5 0.9 1.0 1.3Adjusted* HR:
Reduced Risk Increased Risk
Diabetes Treatments& Cancer Mortality
Bowker et al., Diabetes Care, 2006
Retrospective Cohort StudySaskatchewan, Canada, 1991-1999
1.8 2.0
Sulfonylurea Monotherapy1.30
• Time fixed Cox regression analysis
• Metformin use as the reference group
• Insulin add-on as covariate
Insulin Added1.90
*age, sex, Chronic Disease Score
Diabetes Treatments& Cancer MortalityRetrospective Cohort Study
Saskatchewan, Canada, 1991-1999
• Time varying Cox regression analysis
• SU Monotherapy as the reference group
• Insulin add-on dose-risk gradient
Bowker SL et al., Diabetologia, 2010
0.4 0.6 0.8 1Adjusted HR:
Reduced Risk Increased Risk
Metformin Use
No Insulin Ever (ref)
< 3 Rx/year
3 to 11 Rx/year
≥ 12 Rx/year
0.80
2.22
6.40
1.5 2.0 4.0 6.0
3.33
British Columbia
1 National Diabetes Surveillance System, PHAC2 BC Cancer Agency
BC Cancer Agency
Male Female
BC - DM & Cancer Cohort Study
X
X
X
X
11 yrs
8.5 yrs
8.5 yrs
4 yrs
Subjects may be ‘censored’ due to:- death- leaving province- Dec. 31, 2007
1997 2002 2007
WashoutCancer DM
Index PeriodDM: NDSS case def’nnon DM: sex,
aboriginal match
Follow-up Period
1995
BC - DM & Cancer Cohort Study
Table. Baseline characteristics of the diabetes and controls cohorts (1997-2006)
Diabetes cohort (N=306,210) Control cohort (N=292,782)Baseline characteristics n % n %Sex
Male 162,901 53.2% 155,398 53.1%Female 142,583 46.6% 137,384 46.9%
Age, years30-40 31,437 10.3% 31,545 10.8%40-50 55,894 18.3% 55,552 19.0%50-60 69,237 22.6% 67,533 23.1%60-70 72,519 23.7% 68,976 23.6%70+ 77,123 25.2% 69,176 23.6%Mean (SD) 59.3 (14.2) 58.7 (14.1)
Socioeconomic status (in 1997)
1st quintile 64,798 21.2% 57,248 19.6%2nd quintile 59,155 19.3% 55,311 18.9%3rd quintile 52,765 17.2% 55,373 18.9%4th quintile 49,615 16.2% 56,421 19.3%5th quintile 43,534 14.2% 59,345 20.3%Missing 7,613 2.5% 9,084 3.1%
First-nations / ethnicityYes
BC - DM & Cancer Cohort Study
Table. Cancer and mortality incidence by diabetes index date First neoplasm (any site) Deaths (any cause)
Follow-upIncidence
rateIncidence
rateCohort index date N (years) n % (/1000PY) n % (/1000PY)Incident Diabetes
1997-2001 98,838 559,761 8,268 8.4% 14.8 15,355 15.5% 27.41997 20,128 145,585 2,125 10.6% 14.6 4,418 21.9% 30.31998 19,216 125,097 1,868 9.7% 14.9 3,461 18.0% 27.71999 19,674 112,390 1,633 8.3% 14.5 2,984 15.2% 26.62000 19,402 94,628 1,381 7.1% 14.6 2,444 12.6% 25.82001 20,418 82,062 1,261 6.2% 15.4 2,048 10.0% 25.0
2002-2006 93,468 149,323 2,729 2.9% 18.3 4,791 5.1% 32.1Controls*
None 1994-2001 292,782 2,423,291 30,972 10.6% 12.8 42,555 14.5% 17.6
BC - DM & Cancer Cohort Study
Table. First cancer incident rates by site for incident diabetes and controls cohorts
Incident diabetes (N=192,306) Control cohort (N=292,782)Incidence
rateIncidence
rate
First cancer site n % (/1000PY) n % (/1000PY) IRRAny 10,997 5.7% 15.5 30,972 10.6% 12.8 1.21
Colo-rectal 1,402 0.7% 2.0 3,628 1.2% 1.5 1.32Pancreas 537 0.3% 0.8 625 0.2% 0.3 2.94Lung 1,357 0.7% 1.9 4,107 1.4% 1.7 1.13Breast 1,029 0.5% 1.5 3,355 1.1% 1.4 1.05Cervical/Endometrial 385 0.2% 0.5 904 0.3% 0.4 1.46Prostate 1,510 0.8% 2.1 5,444 1.9% 2.2 0.95Thyroid 71 0.0% 0.1 159 0.1% 0.1 1.53
* 709,085 and 2,423,291 years follow-up in the incident diabetes and controls cohorts
Pharmacologic Agents in BCFormulary
Status DurationData for
Study
Metformin ✔(Open) 1980s ✔
Sulfonylureas ✔(Open) 1980s ✔
Glitazones ✔
rosiglitazone ✔ (Spec Auth) 2005 ✔??
pioglitazone ✔ (Spec Auth) 2005 ✔??
Insulins
Human ✔(Open) 1980s ✔
Long-acting Analogs
glargine ✔ (Spec Auth) 08/2007 ✗
detemir ✗ -- ✗
GLP-1 therapies ✗ -- ✗
BC - DM & Cancer Cohort Study
Source: http://www.health.gov.bc.ca/pharmacare
BC - DM & Cancer Cohort Study
Thank you for your
attention…
Advantages of BC Admin Dataset:- Population-based data (minimize selection bias)- Linkable databases on PHN
- BC Cancer Agency data is rich- Efficient use of available data
- Large populations/samples- Historic data
Disadvantages of BC Admin Dataset:- Accuracy of diagnostic codes / billing data- Incomplete information on potential confounders
- e.g., lifestyle behaviours; clinical data- Limited use of new agents of interest (i.e., glargine, GLP-1)
Epidemiologic Studieswith Large BC Admin
Databases
BC Cancer Agency
Male Female
Validity of DM Case Def’nin Large Admin
Databases
Author‘Gold’
standardSens
%Spec
%PPV
% kappaYoudenIndex*
MB Robinson, 1997
Self-Report
75.5 97.8 72.4 0.72 0.73
NS LeBlanc, 1998
Self-report
62.7 99.4 .. .. 0.62
PEI Van Til, 2001 Diabetes registry
89.2 96.6 62.2 0.71 0.86
ON Hux 2002 Physician charts
86.0 97.0 80.0 0.80 0.83
MB Lix 2006 CCHS 2001
79.5 99.3 87.9 0.82 0.79
*Youden’s index = (Sens + Spec) - 1