The results of this study suggest that the mortality rate from breast cancer in the general population of Taiwan increased for all age groups during the period between 1995 and 2006 (Table 1). This observation is contradictory to what has been observed in the USA, where the overall mortality rate from breast cancer has decreased over the past 15 years .
Although there was an increased risk of mortality from breast cancer in the diabetic patients in all age groups compared with the general population, the increase was most remarkable in the group aged 25–54 years (Table 2). If this age group was to be used as a proxy for premenopausal status, then our results seem to be inconsistent with a meta-analysis that showed a stronger association (overall summary RR 1.19, 95% CI 1.15–1.23) among postmenopausal women or among women of postmenopausal age and no significant association for premenopausal women or women of premenopausal age . Our results indicated that the association was independent of menopausal status, although the mortality rate ratios may not be significant in the group aged 55–64 years according to the sensitivity analyses that compared the mortality rates for the diabetic patients with the highest mortality rates for the general population (Table 2). One possible explanation for the discrepancies could be the ethnicities of the study populations, because none of the studies analysed according the menopausal status in the meta-analysis  had enrolled Asian women.
Three previous studies have shown a significantly higher risk of breast cancer mortality in relation to diabetes in the USA , Italy  and Korea . Except for the Italian study, which recruited patients with type 2 diabetes , the studies conducted in the USA  and Korea  did not distinguish between type 1 and type 2 diabetes. However, because a greater proportion of diabetic patients have type 2 diabetes, the diabetic patients in both these studies probably mainly represent type 2 diabetes. In the present study we showed that there was an increased risk of breast cancer mortality irrespective of whether all diabetic patients or only patients with type 2 diabetes were included in the analysis (Table 2). We did not evaluate whether there was an association in patients with type 1 diabetes because of the relatively small number of such cases in our cohort. Previously, a higher risk of breast cancer mortality was not observed in 23,834 insulin-treated patients with type 1 diabetes (based on an age of onset of <30 years) in the UK . Taken together, the increased risk of breast cancer mortality associated with diabetes probably only applies to patients with type 2 diabetes. Although the UK study did not find a higher risk of breast cancer mortality in insulin-treated patients with type 2 diabetes (based on an age of onset of 30–49 years) , such an assertion of a lack of association in patients with type 2 diabetes is questionable for the following reasons: (1) the sample was distorted and not representative of type 2 diabetes because only insulin-treated patients were selected; (2) the diagnosis based on age of onset of diabetes was not reliable; and (3) there were a relatively small number of cases—only 2,122 insulin-treated diabetic women with the asserted diagnosis of type 2 diabetes, and among these, only 17 cases of breast cancer mortality.
The postulated mechanisms linking diabetes and breast cancer are activation of insulin pathway, activation of the insulin-like growth factor pathway and regulation of endogenous sex hormones [2, 22]. Although we were not able to clarify the mechanisms linking diabetes and breast cancer, our results did provide strong evidence for a link in the Taiwanese population.
Increased leptin and decreased adiponectin levels have been observed in individuals with obesity, type 2 diabetes and the metabolic syndrome, which may also contribute to the increased risk of breast cancer, especially the more aggressive cancers [23, 24]. Breast cancer patients with diabetes were more likely to die (RR 1.76 (95% CI 1.23–2.52) in one study ), suggesting that the prognosis for breast cancer was poorer in the diabetic patients than in those without diabetes. We did not have any biochemical data on our patients and were therefore unable to evaluate how biochemical parameters affected the mortality rates. The significantly higher mortality rate ratios observed in the present study or in studies evaluating breast cancer mortality should not be misinterpreted as significantly higher incidence rate ratios.
The strengths of this study include that it is a prospective follow-up study of a large cohort of diabetic patients, the probable completeness of ascertainment of vital status by obtaining the computerised data file for each individual from the National Register of Deaths using their unique identification number, and the exclusion of possible cases of type 1 diabetes to demonstrate a close link with type 2 diabetes. Nevertheless, there are some limitations. We evaluated cancer mortality and not incidence in this study. If diabetes has an effect on the case fatality rate, then estimation of the mortality rate ratio may not properly reflect the incidence rate ratio. There may be an increased detection bias in the diabetic patients because they may have visited their physicians more frequently, rendering a higher probability of detecting the existence of cancers. However, this may only suggest a higher detection rate of early and slowly progressive breast cancer with a better prognosis in the diabetic patients, which probably would have attenuated the mortality rate. Diabetic patients may have taken more medications than those without diabetes, which may have complicated the situation. For example, metformin, a widely used oral glucose-lowering agent, has been shown to afford protection against breast cancer [26, 27]. Therefore, if patients were taking metformin, this would underestimate the risk of breast cancer associated with diabetes. We were not able to evaluate the effects of medications because of a lack of data; the NHI did not allow access to this information and details were not requested in the baseline questionnaire because most patients are unable to correctly name the medications they use. In the calculation of the mortality rates in the diabetic patients, the person-year denominators for those who were alive were calculated from the date of recruitment until the end of the year 2006 without considering censoring, at the date of emigration, those who emigrated (not available for the study). This may have led to an overestimation of the person-year denominators of the mortality rates and underestimated the mortality rates for the diabetic patients. However, the effect of emigration would be expected to be small because the emigration rate for the total population ranged from 0.05 to 0.60% for the female population over the period of the study according to the yearly report of the Ministry of the Interior of Taiwan . Furthermore, the underestimation of the mortality rates in the diabetic patients could only have underestimated the mortality rate ratios. Therefore, the conclusions of the present study would not be affected.
In conclusion, we have demonstrated an increasing trend of mortality from breast cancer in all age groups in Taiwanese women over the 12 year period between 1995 and 2006. Furthermore, the risk of mortality from breast cancer is increased in diabetic patients, and the magnitude of the risk is largest in the younger age group of 25–54 years, in contrast to a more commonly observed association in postmenopausal women in the Western world. Given that the population is ageing and the incidence of type 2 diabetes is increasing, especially in the younger generation in Taiwan , the affect of breast cancer on mortality should warrant public health attention.