To our knowledge, this is the first study to employ a nation-wide representative cohort to examine the increased risk for invasive breast cancer among women in Taiwan who are undergoing treatment with HT. Because this issue has been heavily debated internationally, we must be cautious about potential confounding factors prior to generating any inferences. However, the following arguments provide a warning to individuals concerning the possible risks of HT. First, because the NHIRD collects all prescription information prospectively, we can rule out the possibility of a recall bias concerning intake dosages and different types of prescriptions: E-alone, E+P HT, progesterone only, or a mixed regimen. Second, in the present study, we included all of the patients who were newly diagnosed with invasive breast cancer between 1999 and 2008 from a simple random sample of one million subjects among the insured general population. Because the rate of insured individuals has been consistently above 96% since 1997, we can rule out the possibility of a selection bias. In fact, our current estimate of 63.7 new breast cancer cases per million person-years is very close to the 66.2 cases per million person-years that was calculated from the National Cancer Registry of Taiwan in 2005. Third, to minimise potential confounding factors according to the indication, we limited the analysis to postmenopausal women (55–79 years old of age), and an association between HT and invasive breast cancer risk was still observed, as summarised in . Indeed, this direct effect of age on breast cancer risk corroborates the result obtained in previous studies 
. Fourth, we attempted to control for potential confounding factors associated with the indication by evaluating ex-users who had ceased hormone medication intake for 6 years or more as the reference group. shows that a discontinued exposure to HT significantly decreases the risk for invasive breast cancer among women of both reproductive age and postmenopausal age, and there is a consistent trend that demonstrates a decreased risk that is accompanied by a longer period of discontinued exposure. Moreover, a significant linear dose-response relationship was observed for an increment of 30 defined daily doses among current users in comparison with ex-users who had ceased HT for more than 6 years. Because we minimized the risk for data falsification, we tentatively concluded that in Taiwan, current users of E-alone and/or E+P HT might have an increased risk for invasive breast cancer.
The present findings shown in and corroborate the results of the WHI, which demonstrated that estrogen plus progesterone might be associated with an increased risk for breast cancer. Although this study was not a randomised control trial, it included a cohort of one million random samples that has been shown to be representative of the population in Taiwan. Since the WHI published their major findings in 2002, Taiwanese women have demonstrated a period of reduced prescription hormone intake 
. In fact, due to fear of any challenge from the patient and/or her family, physicians typically do not prescribe hormones to any patient if there is any doubt regarding the possible beneficial or potential harmful effects of hormones 
. Moreover, the National Health Insurance in Taiwan stipulates a three-month upper limit for repeat prescriptions, which provides another constraint on any unnecessary or harmful medications. The frequency of HT prescribing in Taiwan was somewhat expected lower than in the United States due to this HT prescribing policy (Table S1
). As shown in and , a positive association between E+P and the occurrence of invasive breast cancer in women in Taiwan was still observed under the above circumstances, we think that such a positive association is real.
Breast cancer incidence was notably high in Western countries and the rates of all age groups exceeded those from Taiwan, Japan, and Hong Kong (Table S2
). Although the WHI and some studies in Western countries 
did not observe a positive association between breast cancer and an estrogen exposure of less than 5 years, there have been few reports in Asia to test this hypothesis. As Asian women are generally slimmer than women from Western countries, they might be exposed to a relatively high dose of estrogen based on the guideline recommendations. Given the 12 years of follow-up that were available for testing this hypothesis, we were more likely to detect a positive effect between the prescription of E-alone and invasive breast cancer. Under the NHI system in Taiwan, all individuals who receive HT are required to undergo a follow-up by their physicians no longer than three months. These guidelines increase the likelihood of breast cancer detection during an earlier stage of disease rather than during the invasive stage 
. The positive association persisted under the following different settings. When we limited our analysis to postmenopausal women, or when women who have ceased to use E-alone for more than 6 years were used as the reference group, a positive linear dose-response relationship was observed among current users (). Considering evidence obtained in observational studies that were conducted in Western countries, we suspect that this association is real with respect to the prescribed doses administered in women in Taiwan. As the prevalence of HT exposure differs between Taiwan and other countries, however, we would also expect the burden of HT attributable cancer incidence to differ across the countries, which is of potential interest for public health policy makers. This observation warrants further studies to determine whether a differential genetic susceptibility between Asian and Caucasian postmenopausal women and the use of universal precautions.
In this study and others 
, the combined estrogen-progesterone regimen appeared to be a maximally effective type of estrogen in regard to the occurrence of breast cancer beyond which other types of HT regimens had no or less effect. Adding a progesterone to an estrogen therapy was associated with greater increases in breast cancer risk and with longer washout period for reducing the residual effects compared with the use of estrogen alone. Just as importantly, the disparity between the associations became even larger when we limited our analysis to postmenopausal women. We hypothesize that progesterone play an important role on the occurrence of breast cancer especially in whom has relatively lower endogenous estrogen levels. And, further analysis found that women use mixed regimen were no more susceptible to breast cancer than those use E-alone HT either in reproductive or in postmenopausal age group. The evidence we obtained indicated that the particular type of HT, adding a progesterone continuously to a postmenopausal hormone program, will be more likely to increase the occurrence of breast cancer. However, before drawing any conclusions from the data, further studies are warranted to be replicated with a larger sample size.
The present study has some limitations. First, because the identities of the patients were not available in the NHI reimbursement database, we were unable to obtain any histopathology reports to verify the diagnoses. However, because approval for the registration of invasive breast cancer as a catastrophic illness is based on pathology and/or cytologic evidence and is followed by a full waiver of copayment, such a diagnosis is made only after very serious considerations and is generally accurate. The diagnostic accuracy of invasive breast cancer among the NHI data is corroborated by the considerable agreement between the incidence rate calculated herein and that determined by the National Cancer Registry of Taiwan, in which 95% of the breast cancers are accompanied by histopathologic validation. Second, we were unable to contact the patients directly regarding their use of HRT due to the encryption of their identification numbers; therefore, we cannot exclude the possibility that the subjects used additional phytoestrogenic herbs that were not prescribed. However, because the NHI system provides a comprehensive coverage and because the copayment for prescriptions is consistently 50 NT$ (New Taiwan Dollar) (approximately equal to US $1.5), which is generally less than the cost of herbs that are sold in the markets of Taiwan, the likelihood that the subjects purchased a large amount of other phytoestrogen-containing herbs outside of the NHI database is low. Furthermore, because we had previously limited the cohort to women who did not receive prescription Chinese herbal products during the 12 years of observation, the likelihood that the subjects purchased herbs outside of the services of the NHI becomes even lower. Furthermore, when we limited our analysis to postmenopausal women (55 to 79 years of age) who did not require post-coitus emergency contraceptives or birth control pills, the results showed the same trend as that observed among women aged between 20 to 79 years. However, we are unable to rule out the possibility that the occurrence of breast cancer in postmenopausal women was attributed by the use of a contraceptive pill early in life. Third, we were unable to validate the actual ingested dose of the prescribed HT recorded in the database. A large mean cumulative dose indicates that the patient continued to receive the same prescription for a long period and implies that the patient actually consumed the prescribed medication. Even if the patient did not consume all of the prescribed HT pills, the present findings would only underestimate the effect of the consumed HT. Fourth, because the reimbursement data did not include the selection patterns of phytoestrogen-rich foods, the relative weight and reproductive history of the women, we were unable to control for this factor in the model construction. Because the present study included females from a random sampling cohort, we assumed that such confounding factors would not bias the results.
In summary, the present study found that current users of carefully prescribed E-alone or E+P in Taiwan have an increased risk for invasive breast cancer, especially elderly women. The overall risk seems to be less favourable for E+P in comparison to E-alone. Thus, we recommend that precautionary actions be taken toward the prescription of HT for any duration among postmenopausal women in Taiwan, and the current findings should be incorporated into the established guidelines and emerging risks and benefits of these agents.