Assessing changes in the impact of cancer on population survival without considering cause of death.

Abstract

We have previously argued against the calculation of cancer-specific death rates as philosophically undefined and biased. Deaths attributed to cancer during a particular year occur in patients diagnosed during an unknown distribution of past times, so cancer-specific death rates cannot be used to assess changes in the impact of cancer on survival of the population at specific periods of diagnosis. Our goal was to develop and analyze three measures of the impact of cancer on population survival that do not use the attributed cause of death: 1) the age-adjusted proportion of the population diagnosed with cancer in a particular year and projected to be dead of any cause by a particular age; 2) the same measure corrected for population mortality; and 3) the expected years of life lost to a 20-year-old individual because of the possibility of a diagnosis of cancer. Data on all adults diagnosed with any cancer during the period from January 1973 through December 1990 were obtained from the Surveillance, Epidemiology, and End Results (SEER) Program of the National Cancer Institute. The measures were calculated separately for various combined sex and race groups. Three 2-year diagnosis periods spaced 5 years apart were considered: 1975-1976, 1980-1981, and 1985-1986. A statistical model was used to extrapolate survival beyond observation; the same model was used for patients diagnosed in the three time periods to minimize the effect of possible model misspecification on changes. Cancer incidence increased for three of the four sex-race groups; age-adjusted changes from 1975 through 1976 to 1985 through 1986 were +11.5% for white men, +6.9% for white women, +15.1% for black men, and -9.2% for black women. Human immunodeficiency virus (HIV)-related cancers were responsible for an increased cancer incidence at early ages in white men in the latest time period studied. There was a decrease in the incidence of gynecologic cancers at early ages; the decrease was greater among black women (-55.1%) than among white women (-39.3%). Age- and incidence-adjusted 5-year survival increased by 17.9% for white men, 2.3% for white women, and 7.4% for black men and decreased by 14.1% for black women. When the data from 1985 through 1986 were compared with those from 1975 through 1976, the expected number of years lost to a 20-year-old individual because of cancer changed as follows for the various sex-race groups: +1.4% for white men (-4.0% if HIV-related cancers were not included in the calculation), +2.1% for white women, +12.2% for black men, and +8.8% for black women. For white men and women, there has been an increase in both the incidence of and survival following the diagnosis of cancer; the two effects nearly cancel in our measures. The experience of black men and women has worsened because of increasing incidence or decreased survival.