A method for identifying abrupt changes in U.S. cancer mortality trends

Abstract

BACKGROUND. Summaries of trends in cancer rates evaluated over a fixed, short period are informative, but methods that evaluate trends over a longer time period, identifying when changes occur as well as the magnitude of the changes, can provide additional information. METHODS. Cancer mortality trends from 1973-1995 were determined for more than 15 anatomic sites for white and black males and females using weighted piecewise linear regression analysis with a stepwise selection procedure. The dependent variable was the natural logarithm of the annual mortality rate, and the weight was the annual number of cancer deaths. The variability of estimated change points was examined by bootstrapping the residuals from the resulting models. RESULTS. For black males, cancer mortality rates declined in the 1990s due to decreases in lung, esophageal, oral cavity, and prostate cancer rates. However, there was no significant decline for cancer of the colon and rectum. For white males, cancer mortality rates declined in the 1990s due to declines in cancer of the lung and colon/rectum since the mid-1980s and declines in prostate cancer in the 1990s. For black females and white females total cancer mortality rates declined but not significantly. Cancer rates for all sites except the lung declined significantly in the 1990s for white, but not black females due to declining trends for carcinoma of the colon and rectum since the mid-1980s and for breast cancer in the 1990s. CONCLUSIONS. A method for identifying major changes in cancer trends has been developed. Trends for cancer of the breast and colon/rectum indicate that gaps between rates for blacks and whites are widening.