Methodological Issues - Statistics
The NCI monitors a vast array of cancer statistics to assess progress toward the goal of eliminating suffering
and death due to cancer. The most frequently reported data are descriptive statistics on cancer incidence, survival,
and mortality. However, many other statistical methodologies can be applied to cancer data. NCI's goal is to
develop and improve statistical methods that will aid in the study of cancer and its causes and control.
RRSS investigators are conducting studies to:
- use statistical methods to assess the cost effectiveness of ovarian screening;
- link multiple databases to determine the completeness of coverage ;
- explore methods to estimate hazard rates using data on survival of women with breast cancer;
- use spatial analysis to identify and characterize areas with high rates of late-stage breast and colorectal
cancer;
- develop microsimulation models and a stochastic simulation model of screening for prostate cancer.
Registries Funded to Conduct these Studies
Iowa
New Mexico
Seattle (Puget Sound)
Utah
Key Findings
Investigators proposed a method that applies a microsimulator designed to investigate the effect of prostate
specific antigen (PSA) screening on prostate cancer.
Using statistical methodology, investigators determined that, given our knowledge of lead time, the decline
in prostate cancer mortality could not be entirely explained by PSA testing.
Using a decision analysis model, investigators determined that using a lower bound of 4.0 ng/ml was a more
efficient way to identify men with cancer than the age-specific PSA values.
Future Use
The Cancer Intervention and Surveillance Modeling Network (CISNET) (UO1
(Cooperative Agreement)) grew out of initial work in this area. NCI will continue to develop new and innovative
statistical methods for use in all areas of the analyses of cancer data. Two RO1s
and two R29 awards were funded based on initial
work funded in this area.
|
