PT  - JOURNAL ARTICLE
AU  - P W Dickman
AU  - A Auvinen
AU  - E T Voutilainen
AU  - T Hakulinen
TI  - Measuring social class differences in cancer patient survival: is it necessary to control for social class differences in general population mortality? A Finnish population-based study.
AID  - 10.1136/jech.52.11.727
DP  - 1998 Nov 01
TA  - Journal of Epidemiology and Community Health
PG  - 727--734
VI  - 52
IP  - 11
4099  - http://jech.bmj.com/content/52/11/727.short
4100  - http://jech.bmj.com/content/52/11/727.full
SO  - J Epidemiol Community Health1998 Nov 01; 52
AB  - STUDY OBJECTIVES: Estimation of cancer patient survival by social class has been performed using observed, corrected (cause specific), and relative (with expected survival based on the national population) survival rates. Each of these measures are potentially biased and the optimal method is to calculate relative survival rates using social class specific death rates to estimate expected survival. This study determined the degree to which the choice of survival measure affects the estimation of social class differences in cancer patient survival. SETTING AND PARTICIPANTS: All Finnish residents diagnosed with at least one of 10 common malignant neoplasms during the period 1977-1985 were identified from the Finnish Cancer Registry and followed up for deaths to the end of 1992. DESIGN: Survival rates were calculated by site, sex, and age at 5, 10, and 15 years subsequent to diagnosis for each of three measures of survival; relative survival, corrected (cause specific) survival, and relative survival adjusted for social class differences in general mortality. Regression models were fitted to each set of rates for the first five years of follow up. MAIN RESULTS: The degree of variation in relative survival resulting from social class decreased, although did not disappear, after controlling for social class differences in general mortality. The results obtained using corrected survival were close to those obtained using relative survival with a social class correction. The differences between the three measures were largest when the proportion of deaths from other causes was large, for example, in cancers with high survival, among older patients, and for longer follow up times. CONCLUSIONS: Although each of the three measures gave comparable results, it is recommended that relative survival rates are used with expected survival adjusted for social class when studying social class variation in cancer patient survival. If this is not an available option, it is recommended that corrected survival rates are used. Relative survival rates without the social class correction overestimate social class differences and should be used with caution.