Results

The frequency distributions of the sociodemographic characteristics of study subjects at recruitment are shown in table 1. Among 14 397 subjects recruited, 77% were men and 23% were women. In this cohort, men were older and had a higher educational level than women. There were relatively more women residing in metropolitan areas. A total of 6483 men (68.8%) and 130 women (3.9%) were ever-smokers. Among them, 1258 men (19.4%) and 11 women (8.5%) were ex-smokers. The prevalences of current cigarette smokers in men and women of this study cohort were similar to those of the general population in Taiwan15 as table 2shows.

The characteristics of cigarette smoking among current smokers at recruitment are shown in table 3. The mean duration of smoking was 33 years. The average age of smoking initiation for men and women was 22.4 and 31.5 years, respectively. In general, female smokers had smoked fewer cigarettes per day for a shorter duration than male smokers.

During the period of follow up, 2552 deaths occurred among cohort members. Table 4 shows RRs of dying from various causes of death for current cigarette smokers compared with never-smokers. RRs for some causes of death were not able to be estimated due to the small number of deaths.

In men, cigarette smoking was significantly associated with an increased risk of dying from all causes combined (RR = 1.3), all cancer sites combined (RR = 1.5), stomach cancer (RR = 1.9), liver cancer (RR = 2.2), lung cancer (RR = 3.7), ischaemic heart disease (RR = 1.8), other heart diseases (RR = 1.4), and chronic obstructive pulmonary disease (RR = 1.9). Increased risks that were not statistically significant were noted for cancer of the head and neck (excluding the nasopharynx) (RR = 4.2), nasopharyngeal cancer (RR = 3.9), arterial diseases (RR = 3.0), and peptic ulcer (RR = 4.5). However, cancers of the urinary bladder, pancreas, and oesophagus were not associated with current cigarette smoking.

In women, cigarette smoking was significantly associated with an increased risk of dying from all causes combined (RR = 1.8), lung cancer (RR = 3.6), and peptic ulcer (RR = 17.8). Increased risks that were not statistically significant were noted for all cancer sites combined (RR = 1.7), cancer of cervix uteri (RR = 5.3), ischaemic heart disease (RR = 1.9), and other heart disease (RR = 1.5).

Tables 5-8 show the dose-response relationships between mortality from various causes of death and the quantity of smoking, duration of smoking, age of smoking initiation, and cumulative cigarette smoking in pack-years. After adjustment for age and sex and other potential confounders, statistically significant dose-response relationships were observed between mortality from all causes combined, all cancer sites combined, liver cancer, lung cancer, and ischaemic heart disease and all of these measures of smoking (quantity, duration, age of initiation, and cumulative smoking). There were significant dose-response relationships between oral cancer and smoking quantity, stomach cancer and duration and age of initiation, and chronic obstructive pulmonary disease and age of initiation. We also observed significant dose-response relationships between cancers of the oral cavity and stomach and cumulative cigarette smoking. Those who smoked more than one pack of cigarettes a day (compared with never-smokers) had a 65% increase in risk of dying from all causes of death, 140% from all cancer sites combined, 650% from oral cancer, and 730% from lung cancer, as shown in table 5.

Percentages of deaths from various causes attributable to cigarette smoking in the general population (PAR%) are shown in table 9. In men the PAR% ranged from 21.4% for other heart diseases to 65.4% for peptic ulcer, and in women from 2.1% for other heart diseases to 41.4% for peptic ulcer. The total numbers of deaths in 1994 from various causes among the male and female populations aged 40 years and older in Taiwan are also shown in table 9. Among 58 535 male and 37 410 female total deaths, 8161 (13.9% of total deaths) male and 1216 (3.3% of total deaths) female deaths were attributable to cigarette smoking. Furthermore, 21.3% of the male cancer deaths and 2.9% of the female cancer deaths were attributable to cigarette smoking.

Discussion

This study was the first prospective study to evaluate the health hazards of cigarette smoking in Taiwan. It is important because the epidemic of cigarette smoking-related diseases is emerging in Asia and relatively few prospective studies have been carried out in this region to assess the magnitude of the problem. Cigarette smoking was found to be associated with an increased risk of dying from several major diseases including various cancers, cardiovascular diseases, and pulmonary diseases, similarly to those reported in white populations. However, the numbers of deaths were too small to derive statistically significant RRs for some diseases that have been well documented to be associated with cigarette smoking. These diseases include oral cancer and peptic ulcer in men; and all cancer sites combined, cancer of the cervix uteri, and ischaemic heart disease in women.

Cigarette smoking not only causes excess mortality, but also increases morbidity and may worsen the quality of life. Passive smoking has also been found to cause lung cancer and ischaemic heart disease.17 18 Although the increase in risk associated with passive smoking is smaller than that associated with active smoking, the prevalence of passive smoking is high in Taiwan (roughly equal to the prevalence of smoking among Taiwanese men—50% to 60%). In this study, death rather than disease related to cigarette smoking was chosen as the outcome variable, and health hazards related to passive smoking were not assessed at all. The overall impact of cigarette smoking on the health of people in Taiwan may thus be underestimated in this study. In other words, the RRs, PAR percentages, and smoking-attributable deaths from various causes were estimated in a conservative manner.

Lung cancer mortality increased eightfold during the past 30 years for men and women in Taiwan, which is compatible with the increase in per capita consumption of cigarettes since the second world war. But the prevalence of cigarette smoking remains low in Taiwanese women. Passive smoking is one possible cause of the increase in lung cancer mortality in women. The mortality attributable to passive smoking may be substantial, but was not evaluated in our study.

In comparison with previous prospective studies,2-4 12the risks of dying from various diseases associated with cigarette smoking seemed to be lower in Taiwan than in Western countries. In white populations, cigarette smoking was associated with more than 50% excess risk of dying from all causes combined and approximately 100% excess risk of cancer mortality. In this cohort study, cigarette smoking among men was associated with only 30% excess risk of dying from all causes combined and 50% excess risk in cancer mortality. The corresponding figures for women were 80% and 70%, respectively.

The smoking-associated excess in risk of mortality from all causes combined was 40% for men in Shanghai, China14 and 20% for men and women in Japan.12 In a recent study on cigarette smoking-related mortality in China, the excess in risk of mortality from all causes combined among rural men was only 40%.19 Thus, the risk of dying related to cigarette smoking seems to be lower in Asian populations than in whites. This may be explained by the following reasons: (a) the tobacco epidemic in Asia is at its early stage, and its overall impact on health has not yet emerged; (b) Asian cigarette smokers may start smoking at older ages and may smoke a smaller quantity compared with white smokers; (c) other major risk factors for various smoking-related diseases may be more prevalent in Asian countries, and the RRs for cigarette smoking may thus be underestimated; and (d) the length of the follow-up period in some of the Asian studies was shorter than that of some studies in Western countries. We consider these reasons in more detail below with respect to Taiwan, and Asia more generally.

The prevalence of cigarette smoking in Taiwan has reached its plateau since 1965. In this cohort study, the mean duration of smoking was 33 years with a follow-up period of 12 years. Thus it seems unlikely that the tobacco epidemic in Taiwan is in an early stage.

In this study cohort, the mean cigarette consumption was 17 cigarettes per day, and the mean age of smoking initiation was 22 years. In comparison with smokers in the United States,20 those in Taiwan smoke fewer cigarettes and start smoking at older ages. Thus the lower risks associated with smoking in Taiwan may be due to the older starting age and lower average tobacco consumption compared with white smokers.

As many risk factors other than active cigarette smoking also play important roles in the determination of cancer, cardiovascular diseases, and pulmonary diseases in Taiwan, it is also possible that the RRs estimated for active cigarette smoking may be underestimated. Passive cigarette smoking is one such risk factor. Those who reported themselves as non-smokers in this study were still exposed to environmental tobacco smoke (ETS). In other words, they should be considered as “less exposed” rather than “unexposed” to tobacco smoke. This is especially the case for women than for men. This misclassification of exposure status leads to an underestimate of the RRs for active smoking, with respect to ETS-related conditions such as lung cancer and ischaemic heart disease. This downward bias in RRs for active smoking may be more important in Asian studies than in Western studies because passive smoking is likely to be more prevalent in Asian countries than in Western countries.

Another possible explanation for the lower RRs in Taiwan is the duration of the follow-up period. A longer follow-up period would likely increase the relative risk estimates. The follow-up period was 12 years in our study; eight years in the Japanese study,12 eight years in the Shanghai study,1426 years in the American veterans study,11 20 years in the British doctors study,2 and 16 years in the Swedish men study.7 Thus the duration of the follow-up period was shorter in the Asian studies.

There was no statistically significant increase in smoking-associated risks of dying from cancers of the oesophagus, urinary organs, and pancreas. These cancers have been found to be associated with cigarette smoking in previous studies. The small number of deaths from these conditions during our follow-up period may be one explanation for this discrepancy. Case-control studies to elucidate the risk factors for bladder cancer in Taiwan have consistently shown no association between smoking and bladder cancer; except for cumulative arsenic exposure in the endemic area of blackfoot disease,21 no other risk factors for bladder cancer have been well documented in Taiwan. No studies have been conducted to elucidate risk factors for pancreatic and oesophageal cancer in Taiwan; however, the opposite secular trends of increasing cigarette consumption and decreasing mortality from oesophageal cancer in Taiwan after the early 1960s22suggests that smoking may not be a major risk factor for oesophageal cancer in Taiwan.

In our study, cigarette smoking was found to be associated with an increased liver cancer risk after adjustment for chronic hepatitis B virus carrier status and alcohol drinking. This result is consistent with those reported previously in China,14 Hong Kong,23 and Taiwan.24 Cigarette smoking and chronic hepatitis B virus carrier status were also reported to act synergistically in the development of hepatocellular carcinoma.25 The RR estimate derived by adjustment for chronic hepatitis B virus carrier status and the PAR% of liver cancer deaths for cigarette smoking may thus be underestimated.

The RR of dying from lung cancer was only 3.7 for men in our study. It was similar to that reported from the Japanese cohort study,12 lower than that (RR = 6.5 for ever-smokersvs never-smokers) observed in the Shanghai study,14 and far below those observed in white populations.2-4 However, the RR was 8.3 among current smokers who smoked more than one pack per day, which was compatible with that found in the Shanghai study (RR = 9.4).

Cigarette smoking has been documented as the most important risk factor for obstructive airways disease. However, only a modest increase in the risk of dying from the disease (RR = 1.9) was observed among smokers in this study. The RR was much lower than that (RR ≈ 10 for smokersvs non-smokers) found in Western populations,2 3 but similar to that reported in the Shanghai study (RR = 1.4 for ever-smokers vsnever-smokers).14 One possible explanation for this disparity may be the inaccuracy in diagnosis of chronic obstructive airway disease in Taiwan. Many patients with chronic obstructive pulmonary disease in Taiwan are misdiagnosed as having asthma, pneumonia, or tuberculosis. The misclassification of cases may cause an underestimate of the RR of dying from obstructive airways disease for cigarette smoking. The lower RR may also be due to the later starting age and lower level of cigarette consumption among smokers in Taiwan compared with those in Western countries.

Stroke, after cancer, has been the second leading cause of death in Taiwan since 1982. Although cigarette smoking has been found to be a risk factor for stroke in Western countries, it was not associated with an increased risk of dying from the disease in our study. Hypertension was a more important predictor for stroke in Taiwan, similarly to the report from Shanghai.14

As was shown in reports from Western and Shanghai studies, smoking was an important risk factor for ischaemic heart disease in our study. About 30% of deaths from ischaemic heart disease in Taiwanese men might be attributable to smoking.

In our study, women seemed to have a higher risk of dying from cigarette smoking-related diseases than men. As shown in table 2, female smokers smoked fewer cigarettes, for a shorter duration, and began smoking later in life than did men. But the RRs of dying from various causes among current smokers were higher in women than in men, especially for non-cancer causes of deaths (shown in table 4). Whether this was due to a smaller number of participants and relatively imprecise estimates among women or due to biological and social differences between the sexes deserves further study.

In summary, there was a striking impact of cigarette smoking on the health of Taiwanese. We estimate that 8161 deaths in men aged 40 and older (13.9% of total deaths) were attributable to cigarette smoking in Taiwan in 1994. If all deaths related to cigarette smoking were categorised as a classification code for cause of death, then cigarette smoking would become the second leading cause of death in men in Taiwan, after all cancer sites combined. Tobacco control should be established as the top priority among public health programmes in Taiwan.