2. Study Summary
Research question: Is tobacco smoking associated
with breast cancer?
Study Objective: The relative risk between active
cigarette smoking and breast cancer among the
prospective cohort of Canadian women.
Primary Exposure & Outcome of interest:
• Cigarette smoking and Breast cancer
Catsburg, C., Kirsh, V., Soskolne, C., Krieger, N., & Rohan, T. (2014).
3. Study Summary contn’d
Study Design:
• It was a prospective Case-cohort study
• There were a total of 1,096 cases and 3,314
subcohort participants
Sampling Method:
• Use of self-administered questionnaire
• Random selection of fit
• Smoking variables considered (status, duration,
intensity, age, and quantity)
• Smoking status (never, ever, former, current)
Catsburg et al. (2014)
4. Study Summary contn’d
•Smoking duration (none, <10-<40 or 40+, & yrs prior to
1st full term pregnancy)
• Smoking intensity (none, sticks/day <5-<20, 20+)
• Total cigarettes (none, packs/year <10-<30, 30+)
• Age (never, start/cease smoking <15-<25, 25+)
Catsburg et al. (2014)
5. Study summary contn’d
Statistical analysis:
• Used Cox regression models by Langholz and Jiao
• Models were adjusted for breast cancer and smoking
risk factors
• Sensitivity analysis
• Hazard Ratio (RR) was used to measure association
• 95% confidence interval was used
• Risk factors adjusted for (birth cohort, age at menarche,
use of oral contraceptive, use of hormone therapy,
reproductive history, family history of breast cancer)
Catsburg et al. (2014)
6. Study summary contn’d
Primary measures of association reported:
• The relationship between cigarette smoking, hormone
therapy, age factor, reproductive factors, and breast
cancer
Potential confounders:
• Alcohol use, mammography screening, age of
participants
Catsburg et al. (2014)
7. Study summary cont’d
• Confounding was controlled by stratification
and restriction except environmental tobacco
Catsburg et al. (2014)
8. Study summary cont’d
Summary of major Study results
• Out of 3314 subcohort participants 141 developed
breast cancer
• For adjusted current smokers (HR = 1.04, 95% CI =
0.78 – 1.39); unadjusted (HR = 1.05, 95% CI = 0.79 –
1.38)
• For adjusted former smokers (HR = 1,00, 95% CI =
0.86 – 1.17); unadjusted (HR = 1.01, 95% CI = 0.87 –
1.17)
Catsburg et al. (2014)
9. Critical Analysis
Random error/Selection bias:
• Random error results from inaccurate measurement
related to assessing exposure and outcome; and
mistakes associated with sampling subjects for a study
• It affects the quality of a study (precision)
• Selection bias occurs when an inappropriate criterion is
used to choose study subjects
Aschengrau & Seage, (2014).
10. Critical analysis contn’d
• It leads to an unrepresentative sample of a parent
population by chance
• There was differential loss to follow-up of participants
with respect to exposure and outcome
• Reference subjects exposed to environmental tobacco
may have been included in the study
• Selection bias may be present in the study
Catsburg et al. (2014)
11. Critical analysis contn’d
• Exposure to environmental tobacco can mask
associations
• The magnitude of bias (confounder) may be
significant
• The selection bias may have pulled the observed null
association (RR < 1.2) away from true association
Aschengrau & Seage, (2014)
12. Study limitations:
• Inability to exclude subjects exposed to environmental
tobacco from reference group
• Small sample population for stratified analysis
Catsburg et al. (2014)
13. Generalizability of Study Results
• The study result support current research findings of no
or weak association between exposure and outcome
• The findings reveal no association
• Study did not control for all potential confounder
Catsburg et al. (2014)
14. Application of Study to Public Health
• Future studies may include evaluation of anti-
estrogenic effect on cigarette smoking among
women with NAT2
Public Health Practice Implication of Study:
• Data on investigation between exposure to
environmental tobacco and breast cancer is rare
• More preventable causes of breast cancer can be
identified
Catsburg et al. (2014)
15. References
Aschengrau, A., & Seage, G. R., III. (2014).Essentials of
epidemiology in public health (3rd ed.). Burlington, MA: Jones &
Bartlett.
Catsburg, C., Kirsh, V., Soskolne, C., Kreiger, N., & Rohan, T.
(2014).Active cigarette smoking and the risk of breast cancer: a
cohort study, Cancer Epidemiology, 38(2014), 376-
381.http://dx.doi.org/10.1016/j.canep.2014.05.007
Editor's Notes
Good day everyone, I am Eric Benjamin and an MPH student. My presentation is on study review of the association between breast cancer and active cigarette smoking authored by Catsburg and his Colleagues in 2014.
The use of tobacco has been associated with the development of most types of cancer, including several biological but questionable evidences that it could result to breast cancer. As such, Catsburg and his colleagues investigated the relative risk between active cigarette smoking and breast cancer among the prospective cohort of Canadian women.
A prospective case-cohort study that is considered less prone to bias was conducted. The study was comprised of 1,096 invasive breast cancer cases and 3,314 subcohort subjects drawn from the Canadian Cancer Registry; making up a ratio of propositi to comparison subjects of 0.3 to 3.0 that may be inadequate to detect true association. According to Aschengrau and Seage, (2014), high ratios improve statistical power of a study. Using a self-administered questionnaire, participants were randomly selected according to demographic data – life-style, social characteristics which included age, height, weight, race, education, physical activity, smoking status, and reproductive medical history. These factors have been associated with risk of most cancers.
This slide shows the criterion for which each variable was assessed to obtain bias free results.
The Cox regression analysis also known as Cox Hazard proportion model was used to evaluate the relative risk ratios between the different smoking exposures and breast cancer development. To control for confounding, stratified analysis for birth cohort and smoking were conducted. Stratification for birth cohort included number of live births, age at first live births, whether or not subjects used contraceptives or hormone therapies, whether any family member had suffered from breast cancer, and whether they participated in routine mammography screening. On the other hand, stratified smoking variables included the age at which subjects started smoking prior to first full-term pregnancy, the number of cigarettes smoked in a day, and whether or not subjects never smoked or were current smokers. Lastly, sensitivity analysis was conducted using least exposed participants in each smoking category as reference groups.
Although, Catsburg and his colleagues investigated multiple measures of association between all exposure variables examined and the risk of breast cancer among the prospective cohort of Canadian women, they reported just the relative risk ratio between age, smoking, and reproductive factors and breast cancer. They identified only alcohol use, the age of participants, and mammography screening as factors that could mask the true association between smoking and breast cancer (the primary exposure and outcome of interest).
Apart from the stratified analysis on smoking variables and birth cohort described earlier, the investigators controlled for confounding by restricting routine mammography screening to subjects 50 years of age and above.
Of the 3314 subcohort participants followed in the study, 141 developed breast cancer. However, there was no association between the adjusted variables and the risk of invasive breast cancer among the studied groups.
Despite controlling for potential confounders in the study, information and selection bias were possible in the study. Both selection and information bias may have resulted from the low and slightly differential participation ratio of the case-cohort and subcohort subjects; including missing data on exposure to environmental tobacco. Thus, pulling the observed null association (RR < 1.2) away from true association.
The authors did mention that their inability to assess data for exposure to environmental tobacco was their major weakness of the study.
Given the possibility of selection and information bias, and low level of precision, it would be unwise to generalize the results to any population.
In the study, the authors stated that few research found cigarette smoke to have the capability to prevent endometrial cancer that is associated with estrogen level. Thus, further studies might consider evaluating the anti-estrogenic effect on breast cancer as this may identify preventable causes of cancer.