Related Experiment Videos
Assessing the relationship between marital status and cancer incidence: methodologic considerations
K Zhu1, N S Weiss, S M Schwartz
1Department of Epidemiology, School of Public Health, University of Washington, Seattle 98195.
Cancer Causes & Control : CCC
|January 1, 1994
Summary
Cancer incidence studies comparing marital status found census data led to higher risk estimates than using control cancers. Census-based methods may introduce greater bias in cancer risk assessment.
Area of Science:
- Epidemiology
- Biostatistics
- Cancer Research
Background:
- Registry-based cancer studies often estimate marital status populations using census data or control cancers.
- Discrepancies in these methods can impact the accuracy of relative risk (RR) calculations.
- Previous research highlights the need to validate methodologies in population-based cancer incidence studies.
Purpose of the Study:
- To compare the relative incidence of cancer by marital status using two distinct population estimation methods.
- To evaluate the potential biases introduced by census-based versus control cancer-based population denominators.
- To assess the impact of these methods on cancer risk estimates in diverse racial groups.
Main Methods:
- Utilized cancer registry data from four US populations spanning 1979-1987.
- Calculated age-adjusted relative risks (RR) for selected cancers using census-derived populations.
- Calculated RRs using proportions from other cancers within the same registry as a control.
- Compared results between the two methods, analyzing differences across racial groups and time periods.
Main Results:
- Never-married Black persons showed 1.5 to 2.2 times higher cancer risk with census data compared to married persons; control cancers yielded RRs of 0.7 to 1.1.
- Differences between methods were approximately 20-30% for White populations.
- Bias was more pronounced when extrapolation from census data was required (1981-1987).
Conclusions:
- The census-based method for estimating marital status populations in cancer studies is likely to introduce greater bias than the control cancer method.
- Potential sources of bias include census underenumeration, inconsistent marital status definitions, and extrapolation errors.
- Researchers should carefully consider methodological choices to minimize bias in cancer incidence studies.