Bulletin of the World Health Organization

Trends in smoking and lung cancer mortality in Japan, by birth cohort, 1949–2010

Ikuko Funatogawa a, Takashi Funatogawa b & Eiji Yano a

a. Department of Public Health, Teikyo University Graduate School of Public Health, 2-11-1 Kaga, Itabashi-ku, Tokyo 173-8605, Japan.
b. Clinical Research Planning Department, Chugai Pharmaceutical Company, Tokyo, Japan.

Correspondence to: Ikuko Funatogawa (e-mail: ifunatogawa-tky@umin.ac.jp).

(Submitted: 28 May 2012 – Revised version received: 29 December 2012 – Accepted: 11 January 2013 – Published online: 13 February 2013.)

Bulletin of the World Health Organization 2013;91:332-340. doi: http://dx.doi.org/10.2471/BLT.12.108092

Introduction

Although smoking is known to cause cancer, cardiovascular and respiratory diseases and other health problems,1 almost one billion men and 250 million women in the world smoke.2 Strong evidence of a link between smoking and lung cancer has existed since 19503 and lung cancer mortality is now sometimes used as a measure of a population’s past exposure to smoking.46 Several developed countries, including the United States of America, have experienced high rates of lung cancer mortality in males, with more than 100 deaths per 100 000 men aged 50–54 years of age.7 In Japan, however, this rate of lung cancer mortality is only seen in males in age groups 60–64 or older, even though the prevalence of smoking among Japanese men is high.8,9 Among Japanese women, age-standardized lung cancer mortality rates have increased in the absence of any increase in the prevalence of smoking.10

In general, the probability that you will die of lung cancer at any particular age7,11,12 and the probability that you will smoke are affected by when you were born (i.e. the birth cohort to which you belong).1320 Most of those who die from lung cancer are middle-aged or elderly and die many years after they began smoking. The long time lag and differences between birth cohorts make the full impact of smoking on public health difficult to evaluate. Recent observations from the United Kingdom of Great Britain and Northern Ireland have indicated that age at initiation of smoking may have a marked impact on subsequent lung cancer mortality: the highest levels of lung-cancer-related mortality in middle and old age were observed in the birth cohorts in which more than 63% of males had begun smoking before the age of 20 years.21 Although the early initiation of smoking has been associated with increased health risks in at least two other cohort studies,22,23 the population data needed to explore this link in detail are often unavailable or, if available, have yet to be appropriately analysed.

An annual, cross-sectional nationwide survey on smoking has been conducted in Japan since 1949 by the Japan Tobacco and Salt Public Corporation (known as Japan Tobacco Incorporated from 1985).24,25-28 However, many of the data collected during these surveys – including all of the data collected between 1949 and 1957, the age-specific data collected between 1949 and 1964, and the data on age at initiation of smoking that were collected in the 1958 survey – have never been published or analysed in the medical literature. In this study, age-specific prevalences of smoking and the percentages of people who began smoking early in life (e.g. before the age of 20 years) were evaluated, by birth cohort, using the data previously collected in nationwide surveys in Japan. The results were then related to the corresponding age-specific mortalities from lung cancer.

Methods

Lung cancer mortality

Annual data on lung cancer mortality were extracted from vital statistics collected in Japan between 1950 and 2010.7,29 The data were age-standardized using the World Health Organization’s world standard population.30 For the present study, 5-year age groups in the range from 40–44 years to 80–84 years were investigated. In describing birth cohorts (in this and all other analyses), the median age for each age group was considered to be the age of all individuals in that age group. For example, all participants aged 80–84 years in 1950 were considered to be members of the 1868 birth cohort. The age-specific data on mortality cover the 1868–1968 birth cohorts.

Smoking prevalence

Each of the nationwide surveys on smoking prevalence that have been conducted annually in Japan since 1949 has followed a stratified, two-stage, random-sampling method and has included approximately 5000–32 000 members of the Japanese population aged ≥ 20 years.24,2730 Younger individuals have not been surveyed because Japan banned smoking by individuals younger than 20 years in 1900. The details of the survey method varied with year but included face-to-face interviews and self-administered questionnaires. The data collected allowed the annual evaluation of the prevalence of smoking from 1949 to 2010, broken down by sex and by 10-year age groups from 20–29 years to 50–59 years.9,24,31 The data on smoking prevalence covered the 1897–1985 birth cohorts. Each surveyed individual was simply categorized as a smoker (even if he or she only smoked occasionally) or a non-smoker.

Age at initiation of smoking

Each individual investigated in the 1958 nationwide survey on smoking prevalence (2484 men and 2588 women) was asked at what age he or she had begun smoking. For the present study, other data on age at the initiation of smoking were collected from the questionnaires that were distributed during the National Nutrition Surveys that were run in Japan in 2003 and 2004.3234 Each of these surveys covered approximately 5000 households, in 300 study areas, that were selected by a stratified, two-stage, random-sampling method. In these nutrition surveys, which together included 6320 males and 7108 females, age at initiation of smoking was reported by sex and 10-year age groups ranging from 20–29 years to 60–69 years. The data on age at initiation covered the 1893–1979 birth cohorts. The age of smoking initiation was unknown for 33 (0.6%) of the participants in the 1958 survey and 35 (0.3%) of the participants in the 2003–2004 survey. These 68 individuals were therefore excluded from the final analyses. The proportions of individuals who had initiated smoking by particular ages were calculated (as “cumulative smoking initiation”).

Statistical analysis

For the prevalences of smoking, 3-year moving averages were calculated by age group. Changes with calendar year in both age-standardized lung cancer mortality rates and smoking prevalence were investigated. Changes with age in lung cancer mortality, smoking prevalence and initiation of smoking by certain ages were investigated for the 1925 birth cohort. The 1925 cohort (which, nationally, comprised 3.17 million men and 3.44 million women aged 40–45 years in 1967) was selected for detailed investigation because the corresponding data on both lung cancer mortality at an age of 80–85 years and prevalence of smoking at an age of 20–29 years were available and the relevant – but not exactly corresponding – data on age at initiation of smoking had also been recorded (i.e. the data on those aged 30–39 years at the time of the 1958 survey). Finally, changes with birth year, in lung cancer mortality, smoking prevalence and initiation of smoking by certain ages, were also determined. We calculated 95% confidence intervals (CI) for lung cancer mortality and cumulative smoking initiation.

Similar data previously obtained in the United States were used for reference: the prevalences of cigarette smoking among the 1920–1924 birth cohorts of white men and women, as estimated using retrospective questionnaires,13 and lung cancer mortality rates in the 1920–1924 birth cohorts7 and in white men who had never smoked.35

Results

Changes with calendar year

For both men and women, age-standardized lung cancer mortality rates increased between 1950 and 1996 and then slightly decreased between 1996 and 2010 (Fig. 1). The prevalence of smoking among men aged ≥ 20 years ranged from 74% to 85% over the three decades from 1949 but then gradually decreased, falling to 37% by 2010. The corresponding prevalences for women ranged between 10% and 20% over the six decades from 1949 and showed no clear trend, up or down, over this period.

Fig. 1. Age-standardized lung cancer mortality and smoking prevalence, Japan, 1950–2010
Fig. 1. <b>Age-standardized lung cancer mortality and smoking prevalence, Japan, 1950–2010</b>

The 1925 birth cohort

Among the males of the 1925 cohort, smoking initiation before the age of 20 years was uncommon (reported by 16% of the males; 95% CI: 13–20). Most (60%; 95% CI: 56–64) of the males in this cohort had started smoking when aged 20–25 years, and, once they had started, most of the men had smoked for at least three decades (Fig. 2). The lung cancer mortality rate first reached 100 deaths per 100 000 in males who had reached 60–64 years of age (when 104 deaths per 100 000 were recorded; 95% CI: 100–108), or about 40 years after most of the men had started smoking.

Fig. 2. Age-specific lung cancer mortality, smoking prevalence and cumulative smoking initiation (CSIa) in the 1925 birth cohort, Japan
Fig. 2. <b>Age-specific lung cancer mortality, smoking prevalence and cumulative smoking initiation (CSI<sup>a</sup>) in the 1925 birth cohort, Japan</b>
a Percentage of people in the cohort who had started smoking by a certain age. Note: Age at smoking initiation and smoking prevalence are illustrated by the curves on the left and in the centre of the figure, whereas lung cancer mortality is illustrated by the curves on the right. For reference, the corresponding values for the 1920–1924 birth cohorts in the United States of America7,13 and for lung cancer mortality among white men in the United States who had never smoked35 are also presented.

Although few of the females of the 1925 cohort had begun smoking by the age of 30 years – only 0.1% (95% CI: 0.0–0.8) before reaching 20 years of age and only 7% (95% CI: 5–9) by the age of 29 years of age – about 16% of them began to smoke later in life. Among these females, the lung cancer mortality rate only exceeded 100 deaths per 100 000 at an age of 80–84 years of age (when 133 deaths per 100 000 were recorded; 95% CI: 128–137). In contrast, in the United States, the lung cancer mortality rate among white non-smokers was less than 100 deaths per 100 000, even at 80–84 years of age.35 Compared with their female counterparts, males in the 1925 birth cohort were 12 times more likely to have begun smoking by the time they were 29 years of age, 5 times more likely to be smokers when aged 40–49 or 50–59 years, and 1.2 times and 4 to 5 times more likely to have died from lung cancer when aged 40–44 or when older than 59 years, respectively. When compared against the reference data on men in the United States, Japanese men were less likely to have begun smoking by the age of 20 years, more likely to be smokers and less likely to die from lung cancer in middle-age or when fairly (not very) elderly. When compared against the reference data on women in the United States, Japanese women were less likely to have begun smoking by the age of 20 years, less likely to be smokers and less likely to die from lung cancer in middle age or when elderly. The rise in lung-cancer mortality with increasing age had slowed over the last few decades in the United States but not in Japan.

Changes with birth year among men

In men in each birth cohort, the prevalence of smoking was high but the initiation of smoking before the age of 20 years was relatively uncommon (Fig. 3, Fig. 4, Fig. 5). The prevalence of smoking exceeded 60% in at least one age group within every investigated cohort born before the early 1970s but was lower in more recent birth cohorts. In every birth cohort, fewer than 32% (all 95% CIs between 11% and 35%) of the males began smoking before the age of 20 years and fewer than 3% (all 95% CIs between 0% and 4%) began smoking when they were 14–15 years of age. Although lung cancer mortality among males generally increased over the study period, it was always relatively low given the prevalence of smoking.

Fig. 3. Lung cancer mortality among males, by birth year, Japan
Fig. 3. <b>Lung cancer mortality among males, by birth year, Japan</b>
Note: Mortality is shown for nine age groups covering the middle-aged and elderly.
Fig. 4. Smoking prevalence among males, by birth year, Japan
Fig. 4. <b>Smoking prevalence among males, by birth year, Japan</b>
Note: Smoking prevalence is shown for four age groups.
Fig. 5. Age at smoking initiation among males, by birth year, Japan
Fig. 5. Age at smoking initiation among males, by birth year, Japan
Note: Age at smoking initiation is shown as a cumulative percentage, so that each plotted line shows the cumulative smoking initiation, or percentage of the surveyed individuals who had begun smoking by the age indicated.

The men in the more recent birth cohorts showed relatively low smoking prevalences. However, they were more likely to have begun smoking at a young age (i.e. before the age of 20 years). Within most birth cohorts, particularly the more recent ones investigated, a decrease in the prevalence of smoking was seen with increasing age, probably owing to the cessation of smoking by some middle-aged men. Smoking initiation before the age of 20 years was reported by no more than 22% (all 95% CIs between 11% and 25%) of the males in the earlier birth cohorts investigated (i.e. those born by the late 1930s) but was reported by about 30% (all 95% CIs between 25% and 35%) of the men born after the mid-1940s. The effects of the recent increases in the percentage of Japanese who initiate smoking early in life remain unclear, since much of the lung cancer mortality that will occur in the more recent birth cohorts studied, especially among the elderly, has not yet occurred.

Men born in the late 1930s have shown lower lung cancer mortality rates than men born several years earlier or later. The men born in the late 1930s did have slightly lower smoking prevalences than the men born several years before or after them, were generally older when they began smoking than the males in more recent cohorts, and were more likely to have ceased smoking than the men who were born before them. As indicated in Table 1, complex changes in early initiation, age-specific smoking prevalences and lung cancer mortality rates were noted for selected years and age groups.

Changes with birth year among women

Smoking and the early initiation of smoking were relatively uncommon in females in each birth cohort investigated (Fig. 6, Fig. 7, Fig. 8). The prevalence of smoking was less than 24% among the females in each cohort, and fewer than 12% (all 95% CIs between 0% and 13%) of the females in each cohort reported having started smoking before they were 20 years old. The lung cancer mortality rate among females increased with birth year and exceeded 100 deaths per 100 000 when the cohorts born in the 1910s reached 75–79 years of age.

Fig. 6. Lung cancer mortality among females, by birth year, Japan
Fig. 6. <b>Lung cancer mortality among females, by birth year, Japan</b>
Note: Mortality is shown for nine age groups covering the middle-aged and elderly.
Fig. 7. Smoking prevalence among females, by birth year, Japan
Fig. 7. <b>Smoking prevalence among females, by birth year, Japan</b>
Note: Smoking prevalence is shown for four age groups.
Fig. 8. Age at smoking initiation among females, by birth year, Japan
Fig. 8. <b>Age at smoking initiation among females, by birth year, Japan</b>
Note: Age at smoking initiation is shown as a cumulative percentage, so that each plotted line shows the cumulative smoking initiation, or percentage of the surveyed individuals who had begun smoking by the age indicated.

Some differences were seen between birth cohorts. Lung cancer mortality rates among women older than 65 years were relatively high in cohorts born in the late 1910s and early 1920s. These cohorts experienced relatively high prevalences of smoking compared with the cohorts born over the next four decades. Smoking prevalence was especially low (< 15%) among the women born in the late 1930s. Although no more than 2% (all 95% CIs between 0% and 5%) of the females born before the late 1940s started smoking before the age of  20 years, this proportion gradually increased to 11% (all 95% CIs between 9% and 14%) in the most recent birth cohorts investigated. As with male mortality, many of the lung cancer deaths likely to be seen in females in the most recent birth cohorts have yet to occur.

Discussion

Japan is unusual in that national data on smoking prevalence have been collected annually, in a fairly consistent manner, for more than six decades. Such long-term data are essential in attempts to quantify the full health burden linked to smoking, since some of the adverse effects of smoking may take many years to become apparent. Lung cancer, for example, is mostly diagnosed in the middle-aged and elderly, often in individuals who have been smoking for decades.

Data on age-specific smoking prevalence have been collected since about 1950 in several countries,5,36 but most of the resultant data sets, with very few exceptions,17,20 lack the consistency and specificity needed to define birth cohorts. Although data on smoking in childhood and adolescence are especially sparse, this gap in our knowledge can be filled to some extent by the use of retrospective questionnaires that ask adult respondents to state their age when they first smoked.1316,1820 However, such questionnaires, which were used in the present study to estimate cumulative smoking initiation, can only be administered to survivors and their accuracy may be limited by recall bias. In Japan, retrospective questionnaires were used in several studies conducted between 1983 and 1994; the data collected in them were recently used to estimate the age-specific smoking prevalence rates among individuals born between 1900 and 1952.19

The observation that Japanese males tend to begin smoking later in life than males in the United States might explain why Japanese males have lower lung cancer mortality despite having similar or higher prevalences of smoking in adulthood. Japanese males from the 1900–1945 birth cohorts, for example, showed lower lung cancer mortality rates (> 100 deaths per 100 000 were only seen among men aged 60–64 years or older in Japan but were recorded in men 55–59 years of age or younger in the United States),7 similar or higher prevalences of smoking (> 78% in Japan but < 80% among white males in the United States)13 and rarer initiation of smoking before the age of 20 years (< 30% in Japan but > 50% among white males in the United States).13 Smoking among children and adolescents younger than 15 years is extremely rare in Japan (Fig. 3, Fig. 4, Fig. 5) but has been relatively common (> 10%) within the white male population of the United States for several decades13 and is even more common (> 20%) in some other countries.36 Individuals who begin to smoke before the age of 15 years appear to be at greatly increased risk of developing smoking-related illness.22,23

Since duration of smoking is functionally related to age at initiation, the separate effects of these two variables on health are unclear and difficult to distinguish. Furthermore, inter-individual variation in age at smoking initiation tends to be small within a birth cohort. This makes statistical power low and comparisons within a birth cohort difficult. If, however, age at smoking initiation differs substantially between birth cohorts, comparisons based on aggregated data that have been collected in a consistent manner can be informative.37 More comprehensive comparisons of the initiation of smoking by certain ages, smoking prevalence and lung cancer mortality among birth cohorts, both within and between countries, are needed.

When analysed by birth year, the females of the present study showed a decreasing trend in the prevalence of smoking that was gradually replaced by an increasing trend (Fig. 6, Fig. 7, Fig. 8). The peaks in lung cancer mortality seen among the females born in about 1920 are associated with a relatively high prevalence of smoking in this birth cohort (compared with the values recorded for more recent cohorts) and provide support for a causal link. The increases seen over time in age-standardized lung cancer mortality rates may reflect changes, with birth year, in the prevalence of smoking – or in the initiation of smoking by certain ages – that existed many years ago and may therefore conflict with recent trends in the overall prevalence of smoking (Fig. 1). A four-stage model of cigarette consumption and subsequent mortality among men and women has been proposed and widely used.5,12,38 The model includes a three- to four-decade lag between a rise in the prevalence of smoking and a rise in smoking-attributable mortality.

Among the males of the present study, lung cancer mortality was relatively low in the cohorts born in the late 1930s, possibly as the combined result of a relatively low smoking prevalence, relatively late smoking initiation, and smoking cessation. This may explain the slight decrease recently seen in age-standardized lung cancer mortality (Fig. 1). Males in the cohorts born after the 1930s are less likely to smoke in adulthood than the males born in the 1920s but are more likely to have begun smoking before they reached 20 years of age. The full effects of these differences on mortality are currently unclear because the surviving males who were born after the 1930s have yet to become “elderly”. A repeated annual survey would be useful for monitoring long-term gradual changes in the prevalence of smoking by birth cohort.39,40 Although prospective cohort studies allow individual-level inferences rather than the population-level inferences discussed here, it is often difficult to follow representative samples for long periods.

Legislation, cigarette shortages and social norms may explain some of the present findings. The Japanese government banned smoking by minors in 1900 and cigarettes were rationed in Japan around the time of the Second World War. After 1949, women were targeted in advertising produced by the Japan Tobacco and Salt Public Corporation. In 1985, the Japanese market was opened to foreign manufacturers of cigarettes and the sales competition increased. These events may have accelerated the adoption of smoking by adolescents and young women in Japan.

The transitions seen in Japan may have some features in common with those occurring in countries nearby. In China, the prevalence of smoking among women has remained below 5% – and much lower than the corresponding value for men – for several decades.38 Smoking was a very rare habit among young Chinese women but relatively common among elderly women.41 The mean age at smoking initiation, which, for the 1927–1946 birth cohorts, was recorded as being above 20 years for males and above 25 years for females, has been falling in China.41 In a study conducted in the Republic of Korea in the 1980s, the prevalence of smoking among females was also generally low but relatively high among the elderly.42 In the same country, female lung cancer mortality rates appeared to be highest in the cohort born in the 1920s.7

In summary, lung cancer mortality has increased in Japanese men and women. However, the lung cancer mortality rates in Japanese men remain surprisingly low given the high prevalence of smoking, perhaps because Japanese males rarely begin to smoke early in life. Unfortunately, age at initiation of smoking has been decreasing in both sexes in Japan for the last four decades. Since lung cancer is generally observed in the middle-aged or elderly, the full impact of this worrying trend on lung cancer mortality has yet to be observed.


Funding:

This work was supported by the Japan Society for the Promotion of Science via a Grant in Aid for Scientific Research (80407931) awarded to IF.

Competing interests:

None declared.

References

Share