Our study advanced three main results. First, in all three high-income regions, smoking explained up to 50% of sex differences in period life expectancy between age 50 and 85 over the study period 1950–2015. Second, the decline in these sex differences since approximately 1980 is largely driven by smoking-attributable mortality. Third, whereas smoking-attributable mortality is still increasing for many older female cohorts, it is declining for females in the more recent cohorts in the US and Europe, as well as for males in all three regions.
The massive impact of smoking on mortality is in line with previous studies addressing smoking effects on mortality at the population level; it has been found for the United States [14], in European countries [12, 15, 30,31,32,33,34,35], and worldwide [36]. Smoking affects various causes of death, such as various forms of cancer, cardiovascular disease and multifarious diseases of the respiratory tract [37, 38]. Smoking also explains important differences in life expectancy between countries [39]. Finally, the historical trajectories of divergence between life expectancy with and without smoking-attributable mortality that we found are broadly similar to those previously found for specific countries [40].
Insights into the smoking epidemic across cohorts
Our cohort-by-age analysis of high-income regions confirmed the mortality element of the smoking epidemic model [4, 5]: the increase in smoking-attributable mortality started later among females than for males, and resulted in a later peak at a lower level. Without constructing cohort profiles, we would not have been able to trace two additional important regularities of the smoking epidemic. First, while smoking-attributable mortality in older cohorts still increased, a precipitous decline in smoking-attributable mortality took place in recent cohorts at younger ages. Second, smoking-attributable mortality for males versus females converged across cohorts.
The continuous increase in smoking-attributable mortality in older female cohorts remains cause for concern. It is an essential feature of the smoking epidemic, though, that these deaths result from the high smoking prevalence of women decades ago. More encouraging is the decline in smoking-attributable mortality in recent cohorts at younger ages. Also, smoking prevalence has generally come down over the last decades in the studied regions. For example, in Australia smoking prevalence in females aged 15+ came down from 22.0% in 2000 to 12.4% in 2015 (for males, these numbers were 27 and 14.3%, respectively) [41].
In the same vein, current sex-specific smoking prevalence gives some indication of future sex-specific smoking attributable mortality. In 2015 smoking prevalence generally remained higher in males than in females. For example, in 2015 32% of French men smoked, versus 22% of females; in Germany, 25% of males smoked versus 17% of females; while in the US 14% of males smoked versus 12% of females [7]. Consequently, male smoking-attributable mortality is likely to remain (or become again) higher than female smoking-attributable mortality.
The smoking epidemic and the sex gap in life expectancy
In 1950 the sex gap in period e50 ∣ 85 was 2.0–3.5 years. It subsequently grew to 4.3–4.5 years at the maximum around 1970–1980, and then decreased to 1.8–2.5 years in 2015 for the three regions. The rise, stagnation and decline in the sex differences in survival has been described in detail elsewhere [13, 16,17,18]. Smoking behaviour has been found to explain international differences in the life expectancy sex gap [42,43,44]. We here show that across high-income regions, almost no increase in the sex gap would have occurred without smoking-attributable mortality. Smoking-attributable mortality caused almost all the increase and most of the decrease in the sex gap over the study period.
Although the major part of the decline in the sex gap hitherto is caused by the steep drop in smoking-attributable mortality in males, more and more so this is also due to the increases in smoking-attributable mortality in females overall. For all-causes mortality, in contrast, it has been found that a reduction in the male-female life expectancy gap is, for most countries, due to men dying at lower rates, rather than women at higher rates [18, 45].
We suggest that we may not have seen the end of the narrowing in e50 ∣ 85 sex differences in these regions yet. To date, male smoking-attributable mortality generally still exceeds that of females. Meanwhile, trends are downward for males generally, while for females they are upwards for older cohorts. This suggests scope for further narrowing sex differences in e50 ∣ 85 in these regions. However, the extent to which this may happen seems limited because male smoking prevalence generally remains higher nowadays than female smoking prevalence (see above). Of course, smoking-attributable mortality is not the only factor that affects the sex gap, and there is evidence that mortality from some causes other than smoking may currently be widening the gap [16]. Still, smoking-attributable mortality could overwhelm the effect of mortality from other causes on the sex gap. This could happen especially in countries with a high proportion of women taking up smoking some decades ago, where smoking-attributable mortality for men and women could potentially cross over (e.g. U.K., Denmark and the Netherlands) [8], as we have found for the most recent cohorts in the US and Oceania.
Limitations
One clear limitation of our study is the indirect calculation of the smoking-attributable deaths. Such a limitation is unavoidable: comparisons between different methods to estimate smoking-attributable mortality did not reveal a best-practice method [46, 47], and even if good estimates of smoking prevalence are available to potentially directly estimate smoking-attributable mortality, other factors like smoking intensity are often harder to measure and to take into account in direct estimates.
Since the PGW method [23] extrapolates the lung cancer rates of non-smokers from a US study to other countries, there may be a bias in our estimates for those other countries. Also, the PGW is based on study participants that are more likely than the US overall population to be Caucasian and middle class, and to have achieved a relatively high level of education [47]. However, previous analyses have shown that the indirect estimation by PGW resulted in roughly similar outcomes compared to other indirect estimation techniques [30, 31, 48], so we are confident that our results are broadly reliable. Results obtained making a modification to the PGW method proposed by Rostron [49], discussed in [50], are included in the Additional file 1. Making this modification would not have affected our main conclusion.
As a final limitation, our estimates of smoking-attributable mortality were smoothed over ages, which may lead to minor distortions. We do not expect that to be the case here due to relatively regular source data.