Abstract
BACKGROUND: Different methods exist to estimate smoking attributable cancer mortality rates (Peto and Ezzati methods, as examples). However, the smoking attributable estimates using these methods cannot be generalized to all population sub-groups. A simpler method has recently been developed that can be adapted and applied to different population sub-groups. This study assessed cumulative tobacco smoke damage (smoke load)/non-lung cancer mortality associations across time from 1979 to 2003 among all Massachusetts males and ages 30-74 years, using this novel methodology. METHODS: Annual lung cancer death rates were used as smoke load bio-indices, and age-adjusted lung/all other (non-lung) cancer death rates were analyzed with linear regression approach. Non-lung cancer death rates include all cancer deaths excluding lung. Smoking-attributable-fractions (SAFs) for the latest period (year 2003) were estimated as: 1-(estimated unexposed cancer death rate/observed rate). RESULTS: Male lung and non-lung cancer death rates have declined steadily since 1992. Lung and non-lung cancer death rates were tightly and steeply associated across years. The slopes of the associations analyzed were 1.69 (95% confidence interval (CI) 1.35-2.04, r = 0.90), and 1.36 (CI 1.14-1.58, r = 0.94) without detected autocorrelation (Durbin-Watson statistic = 1.8). The lung/non-lung cancer death rate associations suggest that all-sites cancer death rate SAFs in year 2003 were 73% (Sensitivity Range [SR] 61-82%) for all ages and 74% (SR 61-82%) for ages 30-74 years. CONCLUSION: The strong lung/non-lung cancer death rate associations suggest that tobacco smoke load may be responsible for most prematurely fatal cancers at both lung and non-lung sites. The present method estimates are greater than the earlier estimates. Therefore, tobacco control may reduce cancer death rates more than previously noted.