Predicting lung cancer deaths from smoking prevalence data.

Reliable data on lung cancer burden is not available from most developing countries as cancer registration is lacking. In a previously proposed model to estimate lung cancer deaths in those countries using smoking prevalence data, we estimated the current yearly number of lung cancer deaths in Ethiopia as 3356, a figure far above the WHO estimate of 1343 and the GLOBOCAN of 748. Our aim was to further develop and validate our estimation procedure. We included additional data on risk estimates for lung cancer mortality of ex-smokers and an approximation of duration of smoking into our model and reanalysed study results on non-smoker mortality, thus building two improved models. For validation the number of lung cancer deaths in Germany (2006), the UK (2006), Canada (2004), and Utah, USA (2000) were estimated based on all three models and compared to the observed number of deaths in these countries. We found that the refined model with a modified estimate of lung cancer mortality rates in non-smokers and a more detailed incorporation of smoking dose categories estimates rather well the observed lung cancer deaths in the above countries. With this model, the updated estimate of yearly lung cancer deaths in Ethiopia is 2946 deaths, close to the previous reported estimate. If Ethiopian lung cancer mortality rates in never-smokers and smoking relative risks are the same as in industrialised countries, our models suggests that WHO lung cancer deaths may be underestimated in Ethiopia.