Chapter 2 30 indirect method based on lung cancer mortality as used in this study assume a simplified time pattern of smoking effects on diseases, and reviews comparing direct and indirect methods did not identify a best-practice method.36, 37 The PGW method assumes that the lag between smoking initiation and mortality is the same for lung cancer as it is for other smoking-attributable causes. For lung related diseases this indeed seems to be the case, while for cardiovascular diseases, excess mortality occurs earlier in life as it has a shorter lag period then lung cancer.5 The risk of lung cancer also declines faster upon smoking cessation compared to other lung related diseases.38 Oza et al. 39 showed minimal differences between methods incorporating specific lag times and those that did not. Furthermore, the PGW method assumes that the distribution of lung cancer related deaths among non-smokers is stable across countries and over time, which is not always the case.40 Despite these concerns, several studies have made comparisons of different indirect estimation techniques yielding similar results, showing the robustness of the PGW method.17, 24, 41, 42 This study comes with some limitations. First, not only do smoking rates change over time, educational stratification does as well. Therefore we risk lagged selection bias, i.e. there is a risk that a lower or higher educated person dying in 2000 represents another social structure than someone in 2020.43 This means that we should interpret increasing inequalities in partial life expectancy as in part actual increases, and in part a reflection of changing educational structures.44 On the other hand, the results in our study are similar to those using income quintiles as proxy of socioeconomic status, and quintiles have the advantage of being free of distributional change.17 Furthermore, at older ages information on mortality by educational level was unavailable for some countries, therefore we used partial life expectancy up to age 80. The impact of this choice however is expected to be limited, as previous studies have shown sharp declines in the percentage of smoking-related deaths after age 80.45, 46 In other words, smoking mainly contributes to premature mortality before age 80. Finally, the data coverage differed between countries, especially for Italy and Spain which did not have national-level datasets for long time series, and were instead represented by Turin and Barcelona respectively. However, previous literature has shown that smoking-related mortality in Spain nationally between 2016 and 2019 showed similar educational patterns as those shown for Barcelona in the present study.47-49
RkJQdWJsaXNoZXIy MTk4NDMw