A new method for estimating smoking-attributable mortality in high-income countries

Int J Epidemiol. 2010 Apr;39(2):430-8. doi: 10.1093/ije/dyp360. Epub 2009 Dec 23.


Background: Cigarette smoking is responsible for a massive loss of life in both developed and developing countries. This article develops an alternative to the Peto-Lopez method for estimating the number or fraction of smoking-attributable deaths in high-income countries.

Methods: We use lung cancer death rates as an indicator of the damage caused by smoking. Using administrative data for the population aged > or =50 years from 20 high-income countries in the period from 1950 to 2006, we estimate a negative binomial regression model that predicts mortality from causes other than lung cancer as a function of lung cancer mortality and other variables. Using this regression model, we estimate smoking-attributable deaths based on the difference between observed death rates from lung cancer and expected rates among non-smokers.

Results: Combining the estimated number of excess deaths from lung cancer with those from other causes, we find that among males in 1955 the smoking-attributable fraction was highest in Finland (18%); among women, no country exceeded 1%. By 2003, Hungary had the highest fraction of smoking-attributable deaths among males (32%), whereas the USA held that position among women (24%). Our estimates are remarkably similar to those produced by the Peto-Lopez method, a result that supports the validity of each approach.

Conclusions: We provide a simple and straightforward method for estimating the proportion of deaths attributable to smoking in high-income countries. Our results demonstrate that smoking has played a central role in levels, trends and international differences in mortality over the past half century.

Publication types

  • Research Support, N.I.H., Extramural
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Aged
  • Aged, 80 and over
  • Binomial Distribution
  • Data Interpretation, Statistical*
  • Developed Countries / statistics & numerical data
  • Epidemiologic Methods
  • Female
  • Humans
  • Lung Neoplasms / mortality*
  • Male
  • Middle Aged
  • Risk
  • Smoking / mortality*
  • Social Class